1. Introduction
Inflammatory bowel diseases (IBDs) are chronic disorders characterized by the mucosal inflammation of any segment of the gastrointestinal tract, leading to severe symptoms that could impair the patient’s quality of life. These are classified into two entities—Crohn’s disease (CD) and ulcerative colitis (UC)—which differ in clinical, endoscopic, and histopathological aspects, genetic and epidemiological features, and disease course [1,2]. The management of IBDs has evolved extensively in recent decades with the introduction of novel biologic and molecular therapies [3]. However, within the IBD management evolution pathway, the treatment objectives have changed from clinical remission and endoscopic mucosal healing to histologic remission [4,5]. In this context, endoscopy represents a benchmark for IBD diagnosis and disease monitoring, taking advantage of recent and important technological advancements.
Artificial intelligence (AI) has recently been extensively applied in endoscopy in the context of research settings, with the purpose of helping in the identification and characterization of colorectal polyps, increasing the rate of adenoma detection, and properly managing polyps from the perspective of clinical impact and costs [6]. AI is an umbrella term that includes several model types: natural language processing (NLP) data extraction from unstructured raw text with the generation of human language; the machine learning (ML) model, which enables the AI system to learn and improve from provided data and experiences automatically; the deep learning (DL) model, an ML application which trains the AI system through complex algorithms and deep neural networks, enabling the detection of complex patterns; artificial neural networks (ANNs) for image recognition and diagnosis; and convolutional neural networks (CNNs) for the automatic learning of complex patterns from raw images [7].
To date, the availability of AI systems for applications in endoscopy in the IBD setting has been limited [8]. The importance of a precise and reproducible assessment of mucosal healing to determine the activity of IBD and its response to therapy, thus guiding the choice of patient-focused treatment, has provided a research boost to the field of AI applications in the IBD setting. However, the available studies are characterized by heterogeneity in patient characteristics, study design, research methodology, AI systems, and endoscopic techniques, leading to gaps in the road from the research field to implementation in clinical practice and open debates on the actual usefulness of AI in IBD endoscopy.
In this narrative review, therefore, we assess recent advances in the application of AI-based integrated systems to assess and monitor patients with IBD during endoscopy.
2. Materials and Methods
A non-systematic review of the literature relating to the use of AI in endoscopy in the context of IBD, CD, and UC was undertaken. The PubMed and Scopus databases were searched using a combination of keywords such as ulcerative colitis, Crohn’s disease, inflammatory bowel disease, endoscopy, colonoscopy, capsule endoscopy, device-assisted enteroscopy, artificial intelligence, computer-aided detection, deep learning, machine learning, and neural networks. Studies that were published up until 31 December 2024, had their full text, and were in the English language were considered for this narrative review; these were categorized into the four main key applications of AI in the IBD endoscopy setting: (a) the diagnosis of IBD and the differential diagnosis between CD and UC and between IBD and other non-IBD colitis; (b) the assessment of endoscopic IBD severity; (c) the prediction of IBD histologic activity and clinical outcomes; and (d) the monitoring of disease and the detection of dysplasia occurring in IBD.
3. Limitations of Endoscopy and Advantages of AI in IBD
Endoscopy, encompassing colonoscopy—including magnification, image-enhanced, and microscopic advanced endoscopic techniques—video capsule endoscopy (CE), and device-assisted enteroscopy (DAE), is crucial for diagnosis, clinical management, treatment guidance, and disease monitoring in patients with IBD [9]. The availability of novel biological and molecular molecule-based therapies in the most recent decade has led to the improvement in the quality of life of patients with IBD [3]. At present, endoscopic and histologic remission is considered the primary outcome in the IBD research setting and clinical practice for the assessment of therapeutic efficacy (“treat-to-target” concept), besides clinical remission [10]. However, a significant discrepancy between endoscopic and histological remission has been observed using endoscopy with traditional equipment, as well as a high variability in the treatment objectives between endoscopists [11].
Due to this gap, which has a significant clinical impact on IBD management, recent improvements in endoscopic techniques and image-enhanced programs have also been advanced in the IBD setting.
High-definition (HD) endoscopy is currently recommended to assess IBD activity [12], but some concerns have arisen regarding its ability to accurately predict and determine the latter [13,14,15]. Using white-light endoscopy (WLE), histologically active disease was found in 21.6–23.1% of patients with UC in endoscopic remission [16,17]. In a meta-analysis, WLE did not significantly improve the pooled correlation coefficients between the endoscopic and histologic scores [18]. Chromoendoscopy, which enhances the mucosal superficial patterns and vascular networks using various dyes (dye-based chromoendoscopy, DCE) or electronic optical and digital color-filtering programs (virtual chromoendoscopy, VCE) [19], is currently widely available. However, in IBD endoscopy, contrast dyes present several disadvantages, such as (a) staining of non-inflamed tissue more than inflamed tissue by absorptive dyes (indigo carmine); (b) providing uneven mucosal surface coloring; (c) being time-consuming; (d) requiring endoscopists to have a high proficiency; and (e) lacking enhancement of the subepithelial capillary network [19,20,21]. New-generation endoscopes are usually implemented using VCE techniques (narrow-band imaging, NBI; Olympus Medical Systems, Tokyo, Japan; i-scan optical enhancement, OE; Pentax Medical, Tokyo, Japan; blue-light imaging and linked color imaging, BLI/LCI; Fujifilm, Tokyo, Japan) [22], and different VCE-based endoscopic scores have been proposed for the assessment of IBD activity, with a meta-analysis showing a higher accuracy in predicting histologic remission than WLE [18]. Foremost among them is the unique validated “Paddington International virtual ChromoendoScopy Score” (PICaSSO), evaluating the inflammatory-related vascular and mucosal changes in patients with UC [23]. This endoscopic score could be accurately reproduced with NBI and LCI/BLI, showing a good correlation coefficient with five histologic scores (Pearson’s correlation range: 0.77–0.79), and is therefore currently applied to all electronic VCE platforms [24,25]. However, considerable inter-observer variability, leading to the misevaluation of IBD endoscopic activity, limits a standardized endoscopic assessment using VCE.
Other techniques, such as probe-based confocal laser endomicroscopy (pCLE) and endocytoscopy (EC), show promising applications in the IBD setting. pCLE could be a useful tool for the real-time assessment of histologic activity in both UC and CD (“optical biopsy”), enabling differential diagnosis between UC and CD by visualizing the extent of inflammation and the morphology and density of crypts. The assessment of IBD activity using pCLE to visualize inflammation-related characteristics (such as the disruption of crypts with irregular and wider lumens and microvascular alterations) correlates well with histologic findings [26,27,28,29], even after medical treatment [30,31]. Similarly, through the use of an ultra-magnification endoscopic system in direct contact with the target lesion, EC can provide a highly accurate, real-time, in vivo pathological prediction [32]. The EC system score (ECSS), including characteristics linked to vessels and crypts, strongly correlates with histologic activity [33,34]. Moreover, pCLE and EC could enable the assessment of intestinal barrier permeability and the characterization of the inflammatory infiltrate, respectively [35,36]. However, both techniques require extensive diagnostic training for endoscopists, additional costs and time, and the use of intravenous fluorescein injection for pCLE or the application of mucolytic and contrast agents in EC.
In the context of CD, the topical application of fluorescently labeled adalimumab and vedolizumab during endoscopy allows the detection of membrane-bound TNF1 immune cells and α4β7 integrin, respectively, allowing for the prediction of therapeutic response [37,38]. However, although molecular endoscopic imaging (MEI) is under investigation, significant issues include the additional costs and procedural challenges in endoscopic examination.
In these contexts, the integration of AI systems in endoscopy could help IBD endoscopists because of their ability to analyze a large number of endoscopic images in real time, thus increasing the endoscopic diagnostic accuracy, providing an instantly available endoscopist-independent assessment of the mucosal disease activity, decreasing the inter-observer variability, assisting real-time histological evaluation, and reducing the reading time of CE videos. This would lead to the acquisition of more accurate data, better prediction of histologic remission and clinical outcomes, and improved insights for clinical and treatment decision making in IBD [39,40].
4. AI in Endoscopy for IBD and Differential Diagnosis
Evaluating endoscopic features to distinguish IBD from non-IBD colitis—particularly intestinal tuberculosis, which poses a diagnostic challenge in resource-limited settings—and differentiating CD from UC is complex, requiring precise interpretation by experienced clinicians, as well as intra- and inter-observer coherence.
Several studies, albeit retrospective, have been performed on the AI-based re-analysis of real-world endoscopic images to determine AI’s role in IBD diagnosis.
Different deep learning CNN models (Inception V3—Google AI, Mountain View, CA, USA; ResNet 50—Microsoft Research Asia, Beijing, China; VGG 19—Visual Geometry Group, Oxford, UK, and DenseNet 121—Cornell University, Ithaca, NY, USA) have been compared to determine the best prediction model to accurately distinguish UC from non-UC pathologies and inform the Mayo endoscopic score (MES) of disease severity (inactive/mild and moderate/severe), analyzing 8000 labeled endoscopic images from the HyperKvasir database (the largest available multi-class dataset of images and videos from the gastrointestinal tract—Bærum Hospital, Gjettum, Norway). The DenseNet 121 CNN model provided an area under the receiver operating curve (AUROC) of >0.99 and an accuracy of 98.3%. The addition of Gradient-Weighted Class Activation Maps (Grad-CAMs) improved the visual interpretation of the model over heatmaps [41]. In another similar study, evaluating 6000 endoscopic images from the KVASIR benchmark image dataset, the ResNet-50 CNN model achieved a differential diagnosis accuracy of 99.5% for UC on the validation set [42]. Guimarães et al. found no significant improvement in diagnostic accuracy for distinguishing between IBD and non-IBD colitis using CNN compared to endoscopists (70.9% vs. 72.1%). Only after implementation of the ML Gradient-Boosted Decision Tree (GBDT) approach based on five clinical parameters did the diagnostic accuracy significantly improve (76.6%; AUC = 0.838) [43]. Similarly, by implementing the CNN algorithm with the image pre-processing Pytorch framework (Meta AI, Astor Place, NY, USA) and visualizing the DL model through Grad-CAM (using 6617 colonoscopy images), the diagnostic accuracy for differentiating between CD from intestinal Behcet’s disease and tuberculosis was 65.15% for all images and 72.01% for typical images (p = 0.024) [44]. Inexperienced endoscopists could benefit from CNN-based ML in classifying CD and intestinal tuberculosis, as it showed a sensitivity and specificity of 90% and 77%, respectively [45]. The high yield of AI in distinguishing CD from intestinal tuberculosis has also been found in other studies, with a diagnostic accuracy ranging from 70% to 88.2% [46,47]. A novel classification and regression tree (CART) algorithm, incorporating laboratory, imaging, and endoscopic parameters, found that positive interferon-gamma release assays and circular ulcers are suggestive of intestinal tuberculosis, while involvement of ≥4 segments, along with longitudinal and aphthous ulcers, suggests CD. The overall differential diagnostic accuracy rate for distinguishing CD from intestinal tuberculosis was 88.6%. However, this model was trained on a small sample of patients [48].
AI applied in endoscopy also provides high accuracy in differentiating between CD and UC [7]. By training ResNet50 and ResNeXt-101, two different deep CNNs, on 29,414 and 57,330 colonoscopy still images, respectively, obtained from patients with CD and UC and healthy subjects, algorithms were developed to accurately differentiate these entities. The AI models demonstrated higher diagnostic performance than even the most competent endoscopists. The diagnostic accuracy for IBD ranged from 92% to 99.1% (vs. 92.2% for competent endoscopists and 78% for trainee endoscopists) per patient and from 90.4% to 90.9% (vs. 69.9% for competent endoscopists and 59.7% for trainee endoscopists) per image [49,50]. Importantly, the AI-based algorithm improved the diagnostic yield of non-expert endoscopists by 30.7% (per image) [49]. The accuracy in differentiating CD, UC, and healthy subjects was 92.39%, 93.35%, and 98.35%, respectively, compared to 91.70%, 92.39%, and 97.26% for the best-performing clinicians [49,50]. Another large retrospective DL-based (ResNet34/50/101) study of 11,404 IBD images achieved an accuracy of 90.6% for the differential diagnosis between UC and CD on the validation set. The SI CURA (“Soluzioni Innovative per la gestione del paziente e il follow-up terapeutico della Colite UlceRosA”) database was used as the gold-standard comparator [51]. A CAD method trained to specifically analyze the mucosal architecture on pCLE images from 23 patients with CD and 27 patients with UC, along with nine controls, achieved a sensitivity and specificity of 100% (95% CI = 93 to 100 and 95% CI = 66 to 100, respectively) for diagnosing IBD (p < 0.05 versus controls), as well as a 92% sensitivity (95% CI = 75 to 99) and a 91% specificity (95% CI = 72 to 99) for discriminating between patients with UC and those with CD [52].
Several studies have developed AI-based algorithms for IBD diagnosis using small-bowel and colonic CE videos with varying numbers of training images, comparing the results of endoscopists, both experts and fellows. Among them, one study evaluated the role of AI in video CE for UC, and only three studies were prospective. In the only prospective study including UC lesions (483,644 training datasets and 255,377 validating independent datasets from 31 video CE in 22 patients), the use of the DL ResNet50 framework, with a computational performance of 25 frames/s, achieved diagnostic accuracy rates of 99.2% and 98.3% for the training and validation datasets, respectively [53]. This DL model has been proven to be a useful tool for reducing the burden of image interpretation for endoscopists. The other two prospective studies, which included CD lesions, used the DL ResNet50 and AXARO (Augmented Endoscopy, Paris, France) frameworks, with 7744 training images and 470 images per patient from 130 patients, respectively. Applying the ResNet50 framework with a patient-dependent split of images for training, validation, and testing, the diagnostic sensitivity, specificity, and accuracy for CD-related ulcers were 95.7% (CI = 93.4–97.4), 99.8% (CI = 99.2–100), and 98.4% (CI = 97.6–99), respectively, with two expert readers as comparators. In this study, the diagnostic accuracy was equally high for both the small bowel and the colon [54]. The AXARO framework, applied in a prospective multi-center study of patients with suspected CD, achieved a 97.1% reduction in analyzable images and up to a 94% reduction in the reading time (pooled median review time = 3.2 min per patient) compared to fully read capsules. It also demonstrated a sensitivity and specificity of 92–96% and 90–93%, respectively, and an AUC of 0.91–0.94, highlighting its potential as a rapid tool for ruling out IBD in patients undergoing pan-enteric video CE [55]. The reported diagnostic sensitivity, specificity, and accuracy of CD-related lesions from the other retrospective studies assessing different CNN and DL models on video CE images varied from 88.2% to 98%, 89% to 99.9%, and 90.5% to 99%, respectively [56,57,58,59,60,61,62,63,64,65].
A summary of the studies and results on AI-based diagnosis and differential diagnosis in IBD endoscopy is reported in Table 1.
5. AI in Endoscopy for Assessment of IBD Endoscopic Activity
AI systems in IBD endoscopy have the potential to provide objective and reproducible grading of endoscopic activity in patients with IBD, particularly in the UC setting. Existing endoscopic scores for UC objectively grade the disease severity based on the presence of endoscopic findings without reflecting the picture of clinical severity within each endoscopic category. The most commonly used disease activity index for evaluating response to treatment is the MES, which is easy to apply but has the following notable drawbacks: lack of rigorous validation (poor inter- and intra-observer reliability), limited insertion length, inconsistent distinction between mild (MES 1) and moderate (MES 2) friability, and inability to distinguish between superficial and deep ulcers, reflecting only the most severely affected bowel segment [66,67]. Additionally, the Ulcerative Colitis Endoscopic Index of Severity (UCEIS) score suffers from wide inter-observer variability.
Several AI systems (CNN, deep NN, CAD, DL, support vector machine, residual network, class-based high-resolution network, long short-term memory, and visual geometry group) have been tested in the context of the different endoscopic scores for UC (MES, UCEIS, and PICaSSO), primarily using colonoscopy still images, with fewer studies using endoscopic videos [7]. Expert endoscopists or centrally read videos from clinical trials have been used as comparators. However, all but five of these studies were retrospective. The diagnostic accuracy and AUC of these AI models ranged from 86.54% to 94.5% and from 0.94 to 0.98, respectively [41,68,69,70,71,72].
Iacucci et al. trained a CNN algorithm on 1090 WLE images and VCE videos from 283 patients with UC to grade endoscopic remission/activity and predict histological remission/activity against the grading (using UCEIS and PICaSSO) and agreement provided by experts. This computer model accurately detected endoscopic remission according to UCEIS and PICaSSO, with a sensitivity of 72% and 79%, a specificity of 87% and 95%, and an AUROC of 0.85 and 0.94, respectively. The prediction of histologic remission was similar for the two scoring systems (80% and 85%), while the prediction of the risk of flare was similar to that based on the endoscopic scores provided by endoscopists [73]. However, this model was developed using videos recorded with the i-Scan platform (Pentax, Tokyo, Japan), whilst PICaSSO was recently reported to be valid for other VCE platforms [25]. An accurate distinction between UCEIS 0 (normal mucosa) and UCEIS ≥ 1 (active disease) and between UCEIS 0–3 (mild disease activity) and UCEIS ≥ 4 (moderate–severe disease activity) was achieved using an ML algorithm based on a multi-task learning framework, with accuracies of 90% and 98% (κ = 0.90 and κ = 0.96), respectively. The agreement for UCEIS subdomains (vascular pattern, bleeding, and erosion) was also high (κ ≥ 0.80) [74].
To express inflammation on a continuous scale (from 0 to 10) rather than as a categorical scale, thus providing a comprehensive UC inflammation assessment, a novel AI-based UC Endoscopic Gradation Scale (UCEGS) was generated to express UC severity by training a ranking-CNN using comparative information on UC severity from 13,826 pairs of endoscopic images. UCEGS correlates well with the MES 0–2 scores assigned by IBD expert endoscopists (Spearman’s correlation coefficient = 0.89) and shows a high correlation with the continuous values (0 to 10) provided by endoscopists. However, it offers less variability in estimates for mild- and moderate-disease images compared to the assessments made by endoscopists [75].
In other studies, AI-based differentiation between MES 0 (inactive disease) and MES 1–3 (active disease) achieved an accuracy of 94% (AUROC = 0.997), whereas the distinction between MES 0–1 (remission disease) and MES 2–3 (active disease) ranged from 83.7% to 93%, with an AUROC ranging from 0.966 to 0.998 [76,77,78]. Byrne et al. found that the best MES model performance was for severity levels 0 and 3, with specificities of 94.6% and 87.9% and sensitivities of 85.7% and 69.1%, respectively. For the best UCEIS model, performance was best at severity levels 0 and 5, with specificities of 93.9% and 79.1% and sensitivities of 88.2% and 58.6%, respectively. The accuracy for binary DL-based classification was 94% for MES 0–1 vs. MES 2–3 and UCEIS ≤ 3 versus UCEIS > 3 [79].
The high performance of a CNN-based CAD system in distinguishing MES 0 from MES 1–3 and MES 0–1 from MES 2–3 was confirmed in another study, with AUROCs of 0.86 (95% CI = 0.84–87) and 0.98 (95% CI = 0.97–98), respectively [80]. Interestingly, this performance was superior in the rectum compared to the right- and left-sided colon when distinguishing between MES 0 and MES 1–3 (AUROC = 0.92, 0.83, and 0.83, respectively). However, it was lower in the rectum than in the right- and left-sided colon when identifying MES 0–1 from MES 2–3 (AUROC = 0.99, 0.99, and 0.94, respectively). This could be attributed to topical treatment-induced modifications leading the inflamed mucosa to appear “patchy” or with “skip lesions”, making it more difficult to grade MES using CNN in the rectum accurately. The CNN performance was lower in patients receiving topical treatment compared to those who did not (AUROC = 0.89 and 0.96, respectively) [80].
An 89.1% (sensitivity = 82.3%; specificity = 92.2%) accurate differentiation between mucosal healing (MES 0) and MES 1 was achieved by combining DL- and ML-based CAD diagnostic systems, compared to the 83.3% accuracy achieved by trainee endoscopists [81]. This has prognostic importance, as a higher risk of disease relapse was recently observed in patients with MES 1 compared to those with MES 0, despite the fact that mucosal healing is defined as achieving either MES 0 or 1 [82]. The individual discrimination of MES 1, MES 2, and MES 3 in patients with UC was also achieved using a DL-based algorithm, with AUC values of 0.89, 0.86, and 0.96, respectively, and an overall accuracy of 77.2% [83]. Similarly, another DL-based algorithm, developed using 1672 raw videos from 124 patients with UC, predicted the Mayo Clinic Endoscopic Subscore (MCES) with a high degree of accuracy (AUROC = 0.84 for MCES ≥ 1, 0.85 for MCES ≥ 2, and 0.85 for MCES ≥ 3) [84].
The prediction of MES and UCEIS scores was also performed on full-length endoscopy videos prospectively collected from 249 patients with moderate-to-severe UC within a multi-center clinical phase 2 trial of mirikizumab. This was achieved through training a recurrent neural network (RNN) on score features. The RNN-assisted analysis generated a final endoscopic severity score, achieving high inter-rater agreement with human central readers and demonstrating excellent endoscopic accuracy in predicting endoscopic healing. Specifically, it showed a prediction accuracy of 97% for UCEIS and 95.5% for MES in distinguishing MES 0 from all other score levels [85].
However, scoring selected endoscopic images cannot fully reflect the distribution of inflammation across the entire intestine. Thus, Fan et al. developed a novel DL-based automatic scoring system for assessing inflammatory severity across 85 predetermined areas of different colon tracts from each video-based AI analysis. This system showed high accuracy in predicting each bowel segment’s score, with an accuracy of 86.54% for the MES-scored task and up to 90.7% for the UCEIS-scored task. Additionally, it visualized the distribution of intestinal inflammatory activity using a two-dimensional colorized image [86]. Furthermore, since UC endoscopic assessments report only the maximum severity observed, without taking into account the different extents and gradations of disease severity along the entire colon, Stidham et al. performed a post hoc computed vision analysis that spatially mapped the MES on endoscopic videos from the recent UNIFI trial. This trial evaluated the effects of ustekinumab as an induction and maintenance therapy in moderate-to-severe active UC. The analysis generated a cumulative disease score (CDS) that better quantified the mucosal injury and revealed significant correlations with MES. In addition, it proved more accurate in detecting changes following therapy compared to MES due to its ability to capture variations in the cumulative endoscopic disease severity within each MES level, thus requiring 50% fewer participants to estimate an endoscopic improvement between the ustekinumab and placebo arms. Stratification by pretreatment CDS predicted a greater effectiveness of ustekinumab over the placebo (p < 0.0001), with a more pronounced effect in severe disease compared to mild disease (p < 0.0001) [87]. Another AI-based scoring system, the Ulcerative Colitis Severity Classification and Localized Extent (UC-SCALE), was recently developed by Gutierrez Becker et al. [88] using 4326 sigmoidoscopy WLE videos from phase III Etrolizumab clinical trials. The UC-SCALE, which uses a quality filter for selecting readable images, a scoring system for assigning MCES to each frame, and a camera localization algorithm, achieved similar inter-rater agreement between the UC-SCALE and central and local experienced readers (κ ≥ 0.80). The strengths of this AI-based algorithm include its topological representation as a marker of disease severity and the moderate-to-high correlation of the Aggregated Disease Severity Score (ADSS), calculated using UC-SCALE, with several metrics. These include fecal calprotectin (rs = 0.50), C-reactive protein (rs = 0.45), patient-reported outcomes (rs = 0.45 for stool frequency and rs = 0.40 for rectal bleeding), physician global assessment (rs = 0.45), and total Geboes score (GS) (rs = 0.55) (p < 0.0001 for all metrics) [88].
Recently, colonic tissue oxygen saturation (StO2) was proposed as a measurement for endoscopic healing using the hypoxia imaging algorithm (EP-0002; Fujifilm, Tokyo, Japan), trained on 490 images from 100 patients with UC, based on the characteristic hypoxic microenvironment of the inflamed mucosa [89]. Rectal StO2, assessed by hypoxia imaging colonoscopy, significantly correlated with UC activity as evaluated by the Simple Clinical Colitis Activity Index (p < 0.001), as well as with its subscore, reflecting the urgency of defecation (p < 0.001), at a cut-off of 40.5% for both (AUROC = 0.72 and 95% CI = 0.61–0.84, and 0.74 and 95% CI = 0.62–0.87, respectively). Moreover, StO2 showed moderate accuracy in predicting both endoscopic and histologic activity, with an AUROC of 0.79 (95% CI = 0.74–0.84) for MES ≥ 2 and 0.76 (95% CI = 0.71–0.80) for UCEIS ≥ 2 at cut-offs of 45.5% and 47.5%, respectively. For GS ≥ 3, the AUROC was 0.72 (95% CI = 0.66–0.77) at s 45.5% cut-off. There was an inverse relationship between the StO2 values and MES/UCEIS and GS. However, the higher StO2 values recorded on the right side of the colon might have been influenced by the high concentration of bile components, which affected the detection of the spectral difference between oxyhemoglobin and deoxyhemoglobin for StO2 calculation.
In the CD setting, all studies employed a retrospective design. A multi-brand CNN-based algorithm, trained on 6772 images from single- or double-balloon DAE, was able to automatically detect relevant CD lesions, such as ulcers and erosions, with an accuracy of 98.7% and an AUC–precision recall curve of 1. The reported reading time was 293.6 frames per second, making this AI system potentially applicable in real-life DAE settings [90]. Compared to endoscopists, a DL model (EfficientNet-b528 −Google, Mountain View, CA, USA) combined with Grad-CAM architecture, trained on 155 small-bowel DBE still images from 628 patients with CD, achieved high accuracy in detecting ulcers (96.3%; 95% CI = 95.7–96.7%), non-inflammatory stenosis (95.7%; 95% CI = 95.1–96.2%), and inflammatory stenosis (96.7%; 95% CI = 96.2–97.2%). CD ulcers were also graded on a scale from 1 to 3 according to the ulcerated surface, size, and depth of the ulcers, achieving average accuracies of 87.3% (95% CI = 84.6–89.6%), 87.8% (95% CI = 85.0–90.2%), and 85.2% (95% CI = 83.2–87.0%), respectively [91]. Another study applied a combined DL-CNN and long short-term memory system to pCLE images (testing dataset of 780 images with inflammation and 344 control images) and successfully distinguished between normal and inflamed ileocolonic mucosa in patients with CD. This system showed potential for identifying mucosal healing in inactive CD, with a test accuracy of 95.3% and an AUROC of 0.98, along with irregular crypts and tortuous and dilated blood vessels being indicative of inflamed mucosa [92].
A summary of the studies and results assessing AI-based endoscopic activity using IBD endoscopy is reported in Table 2.
6. AI in Endoscopy for Assessment of IBD Histologic Activity and Prediction of Clinical Outcomes
As stated in the Selecting Therapeutic Targets in Inflammatory Bowel Disease (STRIDE-II) initiative of the International Organization for the Study of Inflammatory Bowel Diseases (IOIBD), combined clinical remission and endoscopic healing are required in long-term treatments [10]. However, persistent histologic inflammation beyond endoscopic mucosal healing is associated with an increased risk of clinical recurrence and the onset of dysplasia in the long term, especially in UC [5,93,94]. A study by Bryant et al. reported a 24% rate of histologically persistent inflammation despite endoscopic remission in patients with long-standing UC [13]. Some studies have reported that conventional WLE does not reliably identify histologic inflammation [95]. Advanced imaging endoscopic techniques such as NBI, i-Scan, CLE, and EC have demonstrated high diagnostic yield in predicting histologic severity, but only when used by expert IBD endoscopists [96,97,98,99]. Moreover, the assessment of histologic inflammation is characterized by high inter-observer variability between pathologists. Thus, AI in IBD endoscopy has the potential to standardize the assessment of histologic disease activity and predict clinical outcomes.
Maeda et al. retrospectively evaluated a CAD system (EB-01; Cybernet Systems Co., Ltd., Tokyo, Japan) to predict persistent histologic inflammation in UC, using images (525 for validation, from 187 patients) obtained using EC (520-fold ultra-magnifying endoscopy) and biopsy samples from six colorectal segments of each patient [100]. The diagnostic accuracy was 91% (95% CI = 83–95%), with very high reproducibility (κ = 1). However, the inter-observer consistency of the histologic GS was not assessed, and a central IBD expert pathologist was not involved. A more sophisticated version of this CAD system, the EndoBRAIN-UC system (Cybernet Systems Corp., Tokyo, Japan), was subsequently adopted in real time during ultra-magnifying colonoscopies in 52 patients with UC. It performed similarly to MES 0 for diagnosing histological healing (defined as GS < 3.1), with a sensitivity of 74.2%, a specificity of 93.8%, and an accuracy of 77.5% (vs. 79.2%, 90.6%, and 81.2%, respectively, for MES 0). This CAD model was also able to identify GS < 3.1 more accurately in MES 1 lesions (p = 0.0169) [101].
These figures were confirmed in prospective studies. Beyond the reported high DNN-based diagnostic accuracy (4187 WLE still images from 875 patients with UC) for endoscopic healing according to UCEIS (90.1%, 95% CI = 89.2–90.9) compared to endoscopists (κ = 0.798, 95% CI = 0.780–0.814), Takenaka et al. demonstrated its ability to identify histologic remission with an accuracy of 92.9% (95% CI = 92.1–93.7) and a κ coefficient of 0.859 (95% CI = 0.841–0.875) compared to the biopsy results [102]. In addition, this DNN model (known as deep neural ulcerative colitis, DNUC) predicted the patients’ prognosis. Patients with DNUC-based mucosal activity were at significantly higher risk of a worse prognosis (p < 0.001 vs. patients with mucosal healing), with hazard ratios for the risk of hospitalization, colectomy, steroid use, and clinical relapse (defined as partial MES ≥ 3, C-reactive protein ≥ 3 mg/L, and fecal calprotectin ≥ 250 mg/g) of 48.4, 46.4, 10.2, and 8.8, respectively, which were similar to those determined by expert endoscopists [103]. Subsequently, the same group applied DNUC to full video colonoscopies, confirming its ability to determine UCEIS, compared to centrally evaluated scoring by IBD expert endoscopists (intra-class correlation coefficient of 0.927; 95% CI = 0.915–0.938). Additionally, it accurately predicted histological remission in 81% of cases, with a sensitivity of 97.9% (95% CI = 97–98.5) and a specificity of 94.6% (95% CI = 91.1–96.9). Of note, the discrepancies between DNUC-based and central reader-based UCEIS scores could be attributed to the presence of inflammatory polyps and inadequate bowel preparation [104]. However, the DNN model was trained only to evaluate the presence or absence of histological inflammation, meaning a detailed histological assessment could not be conducted without biopsy specimens.
A good correlation between the AI-based endoscopic scores and histological activity was observed using a CAD-based algorithm that integrated pixel color data from the redness color map along with vascular pattern detection on WL images (Pentax Medical, HOYA Corporation, Tokyo, Japan). The outputted red density (RD) score correlated with the Robarts Histological Index (r = 0.74, p < 0.0001), MES (r = 0.76, p < 0.0001), and UCEIS (r = 0.74, p < 0.0001) [105]. In contrast to other CAD systems that require thousands of images, this RD approach needs less data, as the algorithm can be modulated sequentially during its development. The RD score also showed potential as an independent predictor of disease course during a follow-up period of five years [106]. An RD score cut-off ≥65 indicated a non-significant increase in the composite endpoint of treatment failure, which included mortality, colectomy due to refractory disease, disease flares, hospitalization, and change in treatment (HR = 2.0, 95% CI = 0.8–5.3). However, this endpoint was assessed retrospectively in only 39 patients with UC, and the results of the ongoing PROCEED-UC trial are awaited to confirm the accuracy and predictive value of the RD score in UC. A limitation of RD technology is its inapplicability to moving images and patients with CD due to the irregular distribution of inflammation and the non-dominant scoring system for endoscopic activity assessment in CD.
As known, the extent of changes in the mucosal peri-cryptal vasculature correlates with the degree of inflammation [52]. Single-wavelength endoscopy (SWE) performed using the prototype system EC-760R endoscope and the VP-7000 processor with a BL-7000 light source (Fujifilm, Tokyo, Japan) provided a real-time in vivo investigation of superficial mucosal crypts, peri-cryptal capillaries, and instances of bleeding (depth up to 5–200 mm). A novel CAD model based on non-magnifying SWE imaging, trained on the corresponding non-magnified HD-WLE images (6926 sets, from 112 UC patients), performed better than a CAD model based on WLE imaging for the assessment of histological remission (GS ≤ 2B.0), with a diagnostic accuracy of 83.3% at initial training (vs. 67.5% for CAD-WLE, p < 0.005) and 95.2% for the validation set [107]. Using the same endoscopic technology, Bossuyt et al. obtained a CAD-based diagnostic accuracy for histologic remission of 86%, compared to 74% and 79% using MES and UCEIS, respectively. Moreover, this CAD-based algorithm reached a 0.694 kappa statistic for correlation with the histologic GS, compared to the correlations between MES or UCEIS and GS (κ = 0.514 and 0.586, respectively) [108]. These studies demonstrate that AI systems could potentially support reducing the number of required biopsy samples and enable immediate therapeutic intervention.
Maeda et al. prospectively applied an AI system (EB-03 prototype; Cybernet Systems, Tokyo, Japan) in real time during colonoscopies in 134 patients with UC in clinical remission who were followed up for 12 months to directly predict clinical relapse (defined as partial MES > 2). The patients were categorized into AI-identified active and healing groups (74 and 61 patients, respectively). The clinical relapse rate was significantly higher in the AI-identified active group (28.4% vs. 4.9%; p < 0.001). The prediction of clinical relapse within 12 months was not significantly different between AI and histology following the analysis of biopsy specimens from 802 segments (accuracy of 58.5% vs. 65.2%; p = 0.316). The prediction of persistent histologic inflammation based on AI had a high accuracy of 93.8% [109]. The same group has more recently proposed an alternative real-time AI-based binary classification, which was applied during colonoscopies in 104 patients with UC in clinical remission: the AI-based vascular healing group and the AI-based vascular active group. Clinical relapse was significantly more frequent in the AI-based vascular active group (23.9% vs. 3%; p = 0.01). In patients with MES ≤ 1, the combination of endoscopic remission and vascular healing parameters provided the highest AUROC for predicting clinical relapse, compared to endoscopic remission alone or combined endoscopic and histologic remission (0.70 vs. 0.65 vs. 0.59) [110]. However, we must not forget that the therapeutic interventions that occurred in the follow-up period might have influenced these promising results. A fully automated three-class MES output (0, 1, and 2 or 3) during colonoscopies in 110 patients with UC in clinical remission, using the EB-UC2 AI prototype (Cybernet Systems, Tokyo, Japan) integrated with a 16-layer Visual Geometry Group network as the architectural framework, could stratify the risk of clinical relapse (defined as partial MES > 2) during the 12-month follow-up. The clinical relapse rates in patient groups classified as MES 0 and 1 were 3.2% (95% CI = 0.1–16.7%) and 24.5% (95% CI = 13.3–38.9%) (p < 0.01), whilst they were 16.2% (95% CI = 8.9–26.2%) and 50% (95% CI = 27.2–72.8%) in patients classified as MES 0 or 1 and MES 2 or 3 (p = 0.03). Furthermore, the inter- and intra-observer reproducibility of non-IBD endoscopists was improved (correlation coefficients = 0.84–0.86 with AI vs. 0.64–0.76 without AI, and 0.89 with AI vs. 0.76 without AI, respectively) [111].
A summary of the studies and results on AI-based histologic activity assessment and prediction of clinical outcomes in IBD endoscopy is reported in Table 3.
7. AI in Endoscopy for IBD Surveillance and Assessment of Dysplasia
Endoscopic surveillance in patients with IBD should be mandatory due to the increased risk of developing colorectal cancer (CRC) [112,113], with an exponential trend rate according to IBD duration (1%, 4%, and 14% at 10, 20, and 30 years, respectively, from IBD diagnosis) [94]. Moreover, IBD-associated dysplasia is often difficult to detect and grade due to chronic inflammation, flat morphology, and margins that are not clearly distinguishable from the surrounding mucosa [114]. Regarding the recent implementation of AI systems into endoscopy, leading to improvement in the detection of colorectal lesions, AI may play a role in detecting early-stage IBD-related dysplasia, identifying patients with IBD who should undergo surveillance colonoscopy, and developing appropriate strategies for surveillance. However, the current data are from case reports and small studies only.
In two case reports, the EndoBRAIN and EndoBRAIN-EYE CAD systems (Cybernet Systems, Tokyo, Japan), previously used successfully for the detection and characterization of colorectal polyps, applied during EC (CF-H290ECI; Olympus, Tokyo, Japan) and high-definition endoscopy (CFHQ290ZI; Olympus, Tokyo, Japan) with NBI in two patients with a long-term history of pan-colitis UC successfully detected a colonic neoplastic lesion and a flat lesion with low-grade dysplasia [114,115]. Guerrero Vinsard et al. retrained an original CADe system (CSPDarkNet53, with cross-stage partial networks) for patient-specific IBD, testing it on HD-WLE images of colorectal lesions in non-IBD patients. The system was evaluated using 1266 HD-WLE and 426 DCE still images of histologically proven dysplastic colorectal lesions in the context of mild-to-moderate mucosal inflammation, achieving good performance with HD-WLE images, showing a 96.8% diagnostic accuracy and a 0.85 AUC (against 77.8% and 0.65, respectively, when using DCE images). Interestingly, the IBD-CADe architecture showed a higher sensitivity in detecting lesions ≤ 10 mm compared to those ≥10 mm (93% for ≤5 mm, 91% for 6–10 mm, and 85% for ≥10 mm). Of note, IBD lesions ≥ 10 mm are often pseudopolyps with a mixed morphology or stalks and overlying mucus. Furthermore, IBD-CADe performed better for lesion types Ip, Is, and IIa (Paris classification), whilst IIb or mixed-morphology lesions were more frequently missed. In addition, it was capable of detecting serrated lesions (epithelial changes and adenomas), even if with a lower true-positive rate (85.7%) than for other dysplastic and non-dysplastic lesions (≥90%). Most missed lesions had higher inflammation scores (missing rates of 7.3% for MES 0, 1.5% for MES 1, and 8.7% for MES 2 and 3) [116]. Another DL model (RetinaNet architecture with ResNet-101 backbone, trained on 478 images from 30 IBD patients) classified lesions into “neoplastic” and “non-neoplastic” with a 93.5% and 87.5% sensitivity and an 80.6% specificity for lesion detection and lesion characterization, respectively [117]. The prediction of neoplasia specifically occurring in IBD was also achieved through a deep CNN-based AI system (EfficientNet-B3), producing a binary classification into “adenocarcinoma or high-grade dysplasia” and “low-grade dysplasia or sporadic adenoma/normal mucosa.” Compared to the diagnostic accuracy provided by four experts and three non-expert endoscopists (for 186 test set images: 77.8%, 95% CI = 74.7–80.8, and 75.8%, 95% CI = 72–79.3, respectively), the diagnostic accuracy of the CNN-based dual-classification was higher, at 79% (95% CI = 72.5–84.6) [118]. However, although in this study the diagnosis of colorectal lesions was performed using p53 and Ki-67 immunostaining, a genetic background analysis was not performed; thus, sporadic colorectal neoplasia might have been included. Such AI systems, if further improved, could help endoscopists, mainly non-experts, in identifying colitis-associated dysplasia or CRC, avoiding unnecessary biopsies.
Finally, the integration of text-based electronic medical records (EMRs) with an NLP- based document-level classification, using the automated retrieval console (ARC) software (available as open source software at
8. Other AI Applications in IBD Endoscopy
With the development of digital pathology, AI algorithms are increasingly employed in histopathological assessments on IBD biopsy specimens. Recently, the Paddington International virtual ChromoendoScopy ScOre (PICaSSO) Histologic Remission Index (PHRI), based only on the presence or absence of neutrophils’ infiltration in the lamina propria and epithelium, was developed using AI within a prospective multi-center study evaluating biopsy samples from 307 UC patients [23]. For each biopsy, from each rectum and sigmoid segment, the worst histologic features were scored using the GS, Robarts Histological Index (RHI), Nancy Histological Index (NHI), ECAP (extent, chronicity, activity and plus) score, and Villanacci Simplified Score. The PHRI score showed a high inter-rater agreement among pathologists (intra-class correlation coefficient of 0.84, 95% CI: 0.78 to 0.90) similarly to RHI and NHI, the strongest correlation with the endoscopic activity according to MES, UCEIS, and PICaSSO (p < 0.05), and the highest correlation with the long-term clinical outcomes (hospitalization, colectomy, and changes in medical therapy due to flare-up), as a PHRI of 1 could accurately stratify the risk of adverse outcomes up to a 12-month follow-up. Subsequently, the PHRI score’s determination through a novel CNN-based architecture DL model detecting neutrophils (training set of 138 biopsies) allowed the differentiation between active and quiescent UC, with 78% sensitivity, 91.7% specificity, and 86% accuracy. PHRI could be successfully implemented into AI models.
Subsequently, Iacucci et al. validated the PHRI on 375 digitalized biopsies using a CNN-VGG16 architecture. The AI classifier accurately distinguished mucosal remission from inflammation with an 89% (95% CI = 0.82–0.94) sensitivity and an 85% (95% CI = 0.80–0.89) specificity (compared to 94% and 76% for RHI and 89% and 79% for NHI, respectively), as well as predicted the corresponding endoscopic remission and activity with an AUROC of 79% (95% CI = 0.75–0.83) and 82% (95% CI = 0.78–0.86) for UCEIS and PICaSSO, respectively. Moreover, it predicted disease flare-ups (hospitalization, UC-related surgery, and changes in UC therapy) for up to 12 months, with a better hazard ratio according to AI-assessed PHRI for histologic remission and activity groups (p < 0.001) (4.64 vs. 3.56 according to pathologist-assessed PHRI) [120]. However, the limitations of this AI-based system include its inability to distinguish different UC severity grades and detect the presence of dysplasia, as well as the lack of worldwide availability of digital pathology. Another DL-based histologic score focused on the detection of eosinophils in sigmoid biopsy specimens (88 UC patients with histologically active disease according to GS and RHI) [121], achieving high agreement with pathologists’ eosinophil counts (interclass correlation coefficients = 0.805–0.917). The eosinophil density was not correlated with histologic activity or biologic use (infliximab, adalimumab, or vedolizumab) but with the disease extent (146.2 cells per mm2 for Montreal E3 vs. 88.2 cells per mm2 for Montreal E2; p = 0.005) and corticosteroid use (62.9 cells per mm2 vs. 124.1 cells per mm2 in non-corticosteroid use; p = 0.006). The DL-based quantification of goblet cell mucus area, as mucin depletion represents a histological risk factor for clinical relapse in MES 0–1, on whole slide images of biopsies (114 UC patients in clinical and endoscopic remission) was proposed by Ozaki et al. for the prediction of clinical relapse (defined as partial MES ≥ 3) [122]. The goblet cell ratio (goblet cell mucus area/epithelial cell and goblet cell mucus area) in specimens of the cecum, ascending colon, and rectum in relapsed patients was lower compared to relapse-free patients (p = 0.010, 0.027, and <0.01, respectively) [123].
Despite recent advances in treatment options, including small molecules and new biologic agents, the response rate to therapy remains modest, and a significant number of patients require a change in treatment over time. However, limited evidence is currently available to guide therapeutic choices. The Endo-Omics study (15 CD and 14 UC patients) demonstrated that in vivo CAD quantitative analysis of pCLE images, including abnormalities in vessel tortuosity, crypt morphology, and fluorescein leakage, predicted the response to anti-TNF or anti-integrin α4β7 therapy after 12 to 14 weeks, with an accuracy of 85% and 80% in patients with UC and CD, respectively (AUROC = 0.93 and 0.79). The ex vivo CAD analysis of fluorescein isothiocyanate-labeled infliximab and vedolizumab staining on the biopsy specimens showed that baseline increased binding of labeled biologics could predict the response to therapy, with a 77% accuracy only in patients with UC (AUROC = 83% vs. 58% in patients with CD) [124]. A spatiotemporal ML-based analysis of CE videos from 101 newly diagnosed and treatment-naïve patients with CD, followed up for six months using the TimeSformer computer vision algorithm (Facebook Research, Menlo Park, California, USA), achieved better prediction for the need of biological therapy compared to the Lewis score (human readers’ grading) and fecal calprotectin (AUROC = 0.86, 0.70, and 0.74, respectively) [125].
9. Conclusions and Future Directions
Overall, as outlined in this review, the implementation of AI algorithms in IBD endoscopy offers substantial benefits, which could revolutionize the management and precision of medicine in the context of IBD. These benefits include the following: (1) enhanced diagnostic accuracy, assisting in the detection of subtle mucosal lesions, ulcers, and inflammation which may be missed by human observers during examinations, differentiating between UC and CD and between IBD and non-IBD colitis, and helping non-IBD expert endoscopists; (2) real-time assessments with prompt decisions about management and therapy based on rapid on-site diagnostic outcomes; (3) standardized and objective evaluation of the disease activity, automatically determining scoring systems like MES or UCEIS or producing new AI-based scores, with a final reduction in the intra- and inter-observer variability across endoscopists (and pathologists); (4) improved efficiency, predicting histologic activity based on endoscopic findings, reducing the need for multiple biopsies, and decreasing the workload for endoscopists and pathologists; and (5) enhanced detection of flat and subtle pre-cancerous lesions, which are visually challenging [8]. Moreover, AI could significantly reduce the reader times of CE videos, with a diagnostic accuracy of up to 99.9%, although the risk of missing lesions should still be assessed. These AI-related strengths would also impact IBD clinical trials beyond clinical practice through helping in central reading [126].
In IBD management, the accurate assessment of mucosal inflammation and healing—both endoscopic and histologic—is crucial for guiding therapeutic and surveillance strategies [5,13,14,15,16,17,18]. However, despite the growing application of AI algorithms in digestive endoscopy, their added value in IBD endoscopy remains unclear, as endoscopic IBD-specific AI algorithms are lacking. Most of the evidence regarding AI applications in IBD endoscopy is provided by retrospective, low-quality, or small-sample studies, particularly for assessing the severity of IBD activity, which remains the most investigated endpoint in AI-supported IBD endoscopy [65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92]. In a recent meta-analysis, the AI diagnostic accuracy for mucosal healing in UC had high sensitivity and specificity but a low yield in accurately differentiating severity grades (e.g., grade 0 vs. 1 and grade 2 vs. 3). Moreover, the meta-analysis detected a moderate-to-high heterogeneity between studies [127]. Similarly, due to the availability of only a few prospective studies, the integration of AI algorithms into clinical practice for diagnosing IBD and distinguishing between IBD and non-IBD colitis or between UC and CD is still in its early stages [41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64]. The prediction of histological remission and response to therapy through HD or ultra-HD endoscopy or real-time histology using pCLE or EC technologies based on AI-based histologic scores also lacks substantial evidence in this preliminary stage [101,102,103,104,105,106,107,108,109,110,111]. The accuracy of AI-based detection and characterization of dysplasia in the IBD setting requires further improvements, as its use is limited by the difficulty in differentiating mucosal and microvascular changes caused by inflammation from those due to malignancy [128]. AI algorithms developed for detecting and characterizing colonic neoplasms may be unsuitable for dysplasia/neoplasia in IBD. Several other challenges also need to be addressed, including the quality of input and output images, which can be compromised by bowel preparation, and the variability in the training datasets on which AI performance depends, which affects the efficiency of AI algorithms in the IBD context. Ethical considerations also need to be considered when integrating AI into clinical and endoscopic practices, which require regulatory approval, data protection measures, and patient privacy. Moreover, elevated AI training and workflow changes, rigorous AI testing, approval by regulatory bodies, extensive external validations in real-world clinical settings, multidisciplinary approaches, and randomized studies and meta-analyses are required before deployment in clinical practice to determine whether AI can effectively improve diagnostic accuracy and forecast clinical outcomes during IBD endoscopy. However, the implementation of AI algorithms in IBD endoscopy holds considerable potential for advancing patient-tailored treatment, monitoring, and surveillance strategies, ultimately improving patient outcomes. This is particularly true with the use of multi-modal AI systems that integrate endoscopic imaging from HD and ultra-HD procedures, patient-level data, radiologic images, and genetic and omics data.
Conceptualization, S.G.G.T. and V.A.; methodology, S.G.G.T. and V.A.; data curation, S.G.G.T., G.A.P., M.L.A., G.D., M.P. and C.D.; writing—original draft preparation, S.G.G.T.; writing—review and editing, V.A.; visualization, G.A.P. and M.L.A.; and supervision, V.A. All authors have read and agreed to the published version of the manuscript.
Not applicable.
Not applicable.
No new data were created or analyzed in this study. Data sharing is not applicable to this article.
The authors declare no conflicts of interest.
The following abbreviations are used in this manuscript:
IBD | Inflammatory bowel disease |
CD | Crohn’s disease |
UC | Ulcerative colitis |
GI | Gastrointestinal |
AI | Artificial intelligence |
NLP | Natural language processing |
ML | Machine learning |
DL | Deep learning |
ANN | Artificial neural network |
CNN | Convolutional neural network |
CE | Capsule endoscopy |
DAE | Device-assisted enteroscopy |
HD | High definition |
WLE | White-light endoscopy |
DCE | Dye-based chromoendoscopy |
VCE | Virtual chromoendoscopy |
NBI | Narrow-band imaging |
OE | Optical enhancement |
BLI/LCI | Blue-light imaging and linked color imaging |
CLE | Confocal laser endomicroscopy |
EC | Endocytoscopy |
PICaSSO | Paddington International virtual ChromoendoScopy ScOre |
ECSS | EC system score |
MEI | Molecular endoscopic imaging |
MES | Mayo endoscopic score |
AUROC | Area under the receiver operating curve |
Grad-CAMs | Gradient-Weighted Class Activation Maps |
UCEIS | Ulcerative Colitis Endoscopic Index of Severity |
CAD | Computer-aided detection |
AUC | Area under curve |
UCEGS | UC Endoscopic Gradation Scale |
MCES | Mayo Clinic Endoscopic Subscore |
RNN | Recurrent neural network |
CDS | Cumulative disease score |
UC-SCALE | Ulcerative Colitis Severity Classification and Localized Extent |
ADSS | Aggregated Disease Severity Score |
GS | Geboes score |
IOIBD | International Organization for the Study of Inflammatory Bowel Diseases |
DNUC | Deep neural ulcerative colitis |
RD | Red density |
CRC | Colorectal cancer |
EMRs | Electronic medical records |
ARC | Automated retrieval console |
PHRI | Paddington International virtual ChromoendoScopy ScOre (PICaSSO) Histologic Remission Index |
RHI | Robarts Histological Index |
NHI | Nancy Histological Index |
ECAP | Extent, chronicity, activity, and plus score |
Footnotes
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Studies on artificial intelligence-based diagnosis and differential diagnosis in inflammatory bowel disease endoscopy.
Study | Year of Publication | Study Design | Endoscopic Technique | Artificial Intelligence Platform | N. of pts | Study Endpoints | Results | Comparator |
---|---|---|---|---|---|---|---|---|
Sutton et al. [ | 2022 | Retrospective | WLE | Inception-V3 | N/R | Diagnosing UC vs. non-UC | AUROCs: | One expert and two trainee endoscopists |
Sharma et al. [ | 2023 | Retrospective | WLE | ResNet-50 | N/R | Diagnosing UC, polyps, esophagitis, and healthy colons | Accuracies:
| Kvasir database |
Guimarães et al. [ | 2023 | Retrospective | WLE | DenseNet + GBDT (five clinical parameters) | Training: 444 pts | Differentiating between IBD and infectious | Overall accuracy: | Three expert endoscopists |
Kim et al. [ | 2021 | Retrospective | WLE | ResNet-34 | 211 CD, 299 intestinal BD, and 217 ITB pts | Differentiating between CD and intestinal BD and ITB | AUROC = 0.78–0.86
| Two experienced endoscopists |
Tong et al. [ | 2020 | Retrospective | WLE | CNN using the Phyton framework | 6399 pts | Differentiating between UC, CD, and ITB | AUROCs: | Endoscopists (number and expertise N/R) |
Lu et al. [ | 2023 | Retrospective | WLE | Text-CNN | 875 CD | Differentiating between CD and ITB | Accuracies: | Endoscopists (number and expertise N/R) |
Lu et al. [ | 2021 | Retrospective | WLE | CART model | Training: 84 CD, 84 ITB | Differentiating between CD and ITB | Accuracy = 88.64% | Endoscopists (number and expertise N/R) |
Ruan et al. [ | 2022 | Retrospective Multi-center | WLE | ResNet-50 | Training: 1358 pts | Differentiating between UC, CD, and normal colons | Accuracies: | Five expert and five trainee endoscopists |
Wang et al. [ | 2022 | Retrospective | WLE | ResNeXt-101 | Training: 217 CD pts, 279 UC pts, and 100 healthy controls | Differentiating between CD, UC, and normal colons | Accuracies: | Six endoscopists of different seniorities |
Chierici et al. [ | 2022 | Retrospective | WLE | ResNet-18 | N/R | Differentiating between CD, UC, and normal colons | Matthews correlation coefficient: | Endoscopists (number and expertise N/R) |
Quénéhervé et al. [ | 2019 | Retrospective Single-center | CLE | CAD system | 23 CD, 27 UC pts, and 9 healthy controls | Diagnosing IBD | IBD diagnosis: | N/A |
Higuchi et al. [ | 2022 | Prospective | CE | ResNet-50 | 22 UC pts | Diagnosing UC | Accuracies: | Five well-trained endoscopists |
Majtner et al. [ | 2021 | Prospective | CE | ResNet-50 | 38 pts with suspected or known CD | Diagnosing CD | Accuracies: | Three experienced gastroenterologists |
Brodersen et al. [ | 2023 | Prospective | CE | AXARO® framework | 131 suspected CD | Diagnosing IBD and CD | AUROCs: | Two specialized observers |
Charisis et al. [ | 2016 | Retrospective | CE | Hybrid Adaptive Filtering- Differential Lacunarity analysis | 13 CD pts | Diagnosing CD | Accuracy: | N/R |
Aoki et al. [ | 2019 | Retrospective | CE | CNN based on Single-Shot Multibox Detector | 65 CD pts | Diagnosing CD | AUROC: | Two expert endoscopists |
Klang et al. [ | 2020 | Retrospective | CE | Xception CNN | 49 CD pts | Diagnosing CD | AUROCs: | One experienced endoscopist |
Barash et al. [ | 2021 | Retrospective | CE | Deep Ordinal Ranking model | 49 CD pts | Grading of ulcer severity | Agreement between consensus reading and automatic algorithm = 67%; | Two and three capsule readers (experiments 1 and 2) |
Klang et al. [ | 2021 | Retrospective | CE | EfficientNet-B5 | N/R | Detecting CD strictures | AUROCs: | N/R |
De Maissin et al. [ | 2021 | Retrospective | CE | ResNet-34 | 63 CD pts | Diagnosing IBD vs. non-IBD | Overall precision = 93.7%; | Three IBD experts |
Ferreira et al. [ | 2022 | Retrospective | CE | CNN using Xception model | N/R | Detecting CD ulcers and erosions | Precision = 97.1% | Three CE experts |
Kratter et al. [ | 2022 | Retrospective | CE | EfficientNet-B4 | N/R | Detecting CD ulcers | Average AUROC = 0.99 | Gastroenterology fellows supervised by capsule experts (number N/R) |
Ribeiro et al. [ | 2022 | Retrospective | CE | CNN using Xception model | 124 CD pts | Detecting CD ulcers and erosions | Accuracy = 99.6% | Three CE experts |
Wang et al. [ | 2019 | Retrospective | CE | Second glance detection framework | 1504 pts (1076 ulcers, 428 normal mucosa) | Detecting CD ulcers | AUROC = 0.9469 (vs. 0.9014 Faster-RCNN and 0.8355 SSD-300) | N/R |
WLE: white-light endoscopy; N/R: not reported; UC: ulcerative colitis; AUROC: area under the receiver operating characteristic; GBDT: Gradient-Boosted Decision Tree; IBD: inflammatory bowel disease; CD: Crohn’s disease; BD: Behcet’s disease; ITB: intestinal tuberculosis; CNN: convolutional neural network; CART: classification and regression tree; CLE: confocal laser endomicroscopy; and CE: capsule endoscopy.
Studies on artificial intelligence-based endoscopy for the assessment of endoscopic activity in inflammatory bowel disease.
Study | Year of Publication | Study Design | Endoscopic Technique | Artificial Intelligence Platform | N. of Patients | N. of Images | Study Endpoints | Results | Comparator |
---|---|---|---|---|---|---|---|---|---|
Sutton et al. [ | 2022 | Retrospective | WLE | Inception-V3 | N/R | 851 still images from the HyperKvasir dataset | Distinguishing MES 0–1 (inactive/mild) from 2 to 3 (moderate/severe) in UC | AUROCs: | One expert and two trainee endoscopists |
Higuchi et al. [ | 2022 | Prospective | CE | ResNet-50 | 22 UC pts | Training: 483,644 images | Assessing endoscopic severity in UC along the entire length of the colon | Accuracy validation dataset: | Five well-trained endoscopists |
Barash et al. [ | 2021 | Retrospective | CE | Deep Ordinal Ranking model | 49 CD pts | 7391 CD images; 10,249 normal mucosa images | Grading of ulcer severity in CD | Overall agreement between manual reading and automatic algorithm = 67% | Three capsule readers |
Kim et al. [ | 2023 | Retrospective | WLE | VGG-16 | 492 UC pts | 984 still images | Differentiating MES 0 vs. 1 | F1-score = 0.92 | Three IBD experts and seven fellow doctors |
Wang et al. [ | 2023 | Retrospective | WLE | High-Resolution Network with Class-Balanced Loss | 308 UC pts | 12,163 still images | Assessing endoscopic activity in UC | MES 0 vs. 123: | Three IBD experts |
Polat et al. [ | 2023 | Retrospective | WLE | ResNet-18 ResNet-50 DenseNet-121 | 564 UC pts | 11,276 still images | Assessing endoscopic activity in UC | QWK Mayo subscores = 0.847 (MobileNet-V3-large)—0.854 (ResNet-18) | Two experienced gastroenterologists |
Qi et al. [ | 2023 | Retrospective | HD endoscopy | ViT network | 768 UC pts | 15,120 still images | Predicting MES in UC | AUROCs: | Six expert endoscopists |
Turan et al. [ | 2022 | Retrospective | HD endoscopy | UC-NfNet | N/R | 673 still images from the HyperKvasir dataset | Classifying colonoscopic UC images | Accuracy = 84.91% | Five board-certified endoscopists with <5 years of experience |
Iacucci et al. [ | 2023 | Retrospective | WLE and VCE videos | ResNet-50 | 283 UC pts | 1090 endoscopic videos (67,280 frames) | Distinguishing UC endoscopic remission (ER) | WLE videos: | Experienced endoscopists from the PICaSSO group |
Patel et al. [ | 2022 | Prospective | HD endoscopic videos | Multi-task learning algorithm (MLA) | 73 UC pts | 38,124 frames | Distinguishing UCEIS 0 vs. active disease, UCEIS 0–3 vs. moderate/severe disease | UCEIS 0 vs. ≥1: | Three IBD experts |
Takabayashi et al. [ | 2024 | Retrospective Multi-center | HD endoscopy | Ranking-CNN | 812 UC pts | 13,826 pairs of still images | Grading UC severity by UC Endoscopic Gradation Scale (UCEGS) | Spearman’s correlation coefficients: | Seven IBD expert endoscopists |
Lo et al. [ | 2022 | Retrospective | WLE | Inception Net-V3 | 467 UC pts | 1484 still images | Distinguishing active vs. healed mucosa; differentiating levels of endoscopic disease activity |
| Two IBD experts |
Yao et al. [ | 2021 | Retrospective Multi-center | HD endoscopic videos | Inception-V3 | 157 UC pts | 175 videos | Grading endoscopic UC disease | Informative image classifier: AUROC = 0.93; | Two IBD experts |
Stidham et al. [ | 2019 | Retrospective | HD endoscopy | Inception-V3 | 3082 UC pts | 16,514 still images; | Grading endoscopic UC disease | MES 0–1 vs. MES 2–3: | Two IBD experts |
Byrne et al. [ | 2023 | Prospective | HD endoscopy | EfficientNet-B3 | N/R | 134 videos (1,550,030 frames) | Predicting MES and UCEIS in UC pts | At section level: | One global central reading expert, six gastrointestinal specialists, and twenty gastrointestinal trainees |
Ozawa et al. [ | 2019 | Retrospective | WLE | CNN-based CAD system on GoogLeNet architecture | 841 UC pts | 26,304 still images | Identifying normal mucosa (MES 0) vs. healing state (MES 0–1) | AUROCs MES 0 vs. 1–3: | N/R |
Huang et al. [ | 2021 | Retrospective | HD endoscopy | DNN, support vector machine, k-nearest neighbor network | 54 UC pts | 856 still images | Diagnosing mucosal healing in UC | Accuracies: | Two reviewers |
Bhambhvani et al. [ | 2021 | Retrospective | HD endoscopy | ResNeXt-101 | 777 active UC pts | 777 representative still images from the HyperKvasir dataset | Grading individual MES in UC | AUROCs: | One experienced gastroenterologist and one fellowship physician in gastroenterology |
Gutierrez Becker et al. [ | 2021 | Retrospective | WLE videos from etrolizumab Phase II Eucalyptus and Phase III Hickory and Laurel clinical trials | Quality control model-CNN | 1105 UC pts | 1672 videos | Grading individual MES in UC | AUROCs: | Expert gastroenterologists |
Gottlieb et al. [ | 2023 | Prospective | WLE videos from a phase II trial of mirikizumab | Recurrent neural network | 249 UC pts | 795 videos | Predicting central reader scores | MES: | Expert central readers |
Fan et al. [ | 2023 | Retrospective | WLE | ResNet-50 | 332 UC pts | 5875 still images and 20 full-length videos | Scoring full-length intestinal inflammatory activity | Mayo-scored task:
| Four endoscopists with 30, 11, 4, and 6 years of experience |
Stidham et al. [ | 2024 | Retrospective | WLE videos from the UNIFI clinical trial | Computer vision analysis that spatially mapped MES to generate the cumulative disease score (CDS) | 748 induction and 348 maintenance UC pts | N/R | Quantifying endoscopic severity in UC; CDS vs. MES for differentiating response to ustekinumab vs. placebo | CDS:
| N/R |
Gutierrez Becker et al. [ | 2024 | Retrospective | WLE videos from phase III Etrolizumab clinical trials | QC model-V7 platform | 1953 UC pts | 4326 sigmoidoscopy videos | Evaluating endoscopic severity and disease extent in UC using Ulcerative Colitis Severity Classification and Localized Extent (UC-SCALE) | QWK between UC-SCALE and MCES by central reading: | Central and local reading (leading IBD gastroenterologists) |
Akiyama et al. [ | 2024 | Retrospective | WLE | EP-0002 function by Fujifilm | 100 UC pts | 490 images | Assessing colonic tissue oxygen saturation (StO2) for evaluation of clinical, endoscopic, and histologic activity in UC | Rectal StO2 correlated with Simple Clinical Colitis Activity Index (p < 0.001) | Three board-certified endoscopists and two board-certified pathologists |
Martins et al. [ | 2023 | Retrospective | DAE | XCeption model multi-brand CNN | 250 DAE exams | 6772 images | Detecting ulcers and erosions in CD | Sensitivity = 88.5% | Two experienced endoscopists |
Xie et al. [ | 2024 | Retrospective | DBE | EfficientNet-B5 | 628 pts | 28,155 small-bowel DBE images | Detecting and objectively assessing small-bowel CD | Accuracy: | Two experienced endoscopists |
Udristoiu et al. [ | 2021 | Retrospective | CLE | DL combined with CNN and long short-term memory (LSTM) | 54 UC pts (32 with known active disease, 22 controls) | 6205 images | Distinguishing between normal and inflamed colonic mucosa in CD | Normal colonic mucosa: round crypts | N/R |
WLE: white-light endoscopy; N/R: not reported; UC: ulcerative colitis; AUROC: area under the receiver operating characteristic; IBD: inflammatory bowel disease; CD: Crohn’s disease; MES: Mayo endoscopic score; QWK: quadratic weighted kappa; VCE: virtual chromoendoscopy; UCEIS: Ulcerative Colitis Endoscopic Index of Severity; CNN: convolutional neural network; DL: deep learning; DAE: device-assisted enteroscopy; DBE: double-balloon enteroscopy; CLE: confocal laser endomicroscopy; CE: capsule endoscopy; and AUPRC: area under precision–recall curve.
Studies on artificial intelligence-based endoscopy for assessment of the histologic activity of inflammatory bowel disease and the prediction of clinical outcomes.
Study | Year of Publication | Study Design | Endoscopic Technique | Artificial Intelligence Platform | N. of Patients | N. of Images | Study Endpoints | Results | Comparator |
---|---|---|---|---|---|---|---|---|---|
Iacucci et al. [ | 2023 | Retrospective | WLE and VCE videos | ResNet-50 | 283 UC pts | 1090 endoscopic videos (67,280 frames) | Predicting histology and risk of flare | VCE videos:
| Experienced endoscopists from the PICaSSO group |
Maeda et al. [ | 2019 | Retrospective | EC | CAD system (EB-01) | 187 UC pts | Training: | Predicting persistent histologic inflammation in UC |
| Endoscopists (number and experience N/R) |
Omori et al. [ | 2024 | Retrospective | WLE ultra-magnifying endoscopy vs. conventional light non-magnifying endoscopy | EndoBRAIN-UC system | 52 UC pts | N/R | Diagnosing histologic healing in UC |
| Three endoscopists |
Takenaka et al. [ | 2020 | Prospective | WLE | DNUC (deep neural network for evaluation of UC) | Training: 2012 UC pts | Training: 40,758 still images | Predicting endoscopic and histologic remission |
| Three endoscopists with 11, 13, and 32 years’ experience in IBD-endoscopy |
Takenaka et al. [ | 2021 | Prospective | WLE | DNUC (deep neural network for evaluation of UC) | 875 UC pts | 4187 still images | Predicting UC pts prognosis |
| Three endoscopists with 11, 13, and 32 years’ experience in IBD endoscopy |
Takenaka et al. [ | 2022 | Prospective | WLE videos | DNUC (deep neural network for evaluation of UC) | 770 UC pts | Colonoscopy full videos (number N/R) | Real-time detection of UC histologic mucosal inflammation |
| Two central reviewer endoscopists with 12 and 14 years’ experience |
Bossuyt et al. [ | 2020 | Prospective | WLE with red density (RD) function | CAD RD-based algorithm | 29 UC pts, six healthy controls | Number of images N/R | Determining UC endoscopic and histologic activity | RD correlated (p < 0.0001) with the following:
| Two groups of two IBD endoscopists (two with >10 years’ experience) |
Sinonquel et al. [ | 2023 | Retrospective | WLE with red density (RD) function | CAD RD-based algorithm | 39 UC pts from RD pilot study, 6 healthy controls | Number of images N/R | Predicting sustained clinical remission using RD | RD ≥ 65:
| N/R |
Sinonquel et al. [ | 2024 | Prospective | WLE and SWE | CAD models CNN-based (ResNet-50, VoVNet) | 112 UC pts | 6926 images | Assessing accuracy of WLE-CAD and SWE-CAD systems for UC histologic activity | SWE-CAD: | Number and experience of endoscopists N/R |
Bossuyt et al. [ | 2021 | Prospective Single-center | SWE | CAD model | 58 UC pts | 113 still images | Automatically evaluating changes in mucosal peri-cryptal vascular structures associated with UC activity (number of bleeding pixels, number of pixels with high density) | CAD histologic remission: | Number and experience of endoscopists N/R |
Maeda et al. [ | 2022 | Prospective | EC | Endo-BRAIN-UC | 61 UC pts healing group, 74 UC pts active group | 44,097 images | Stratifying relapse risk of UC pts in clinical remission |
| Two endoscopists trained on the AI system in at least three UC cases |
Kuroki et al. [ | 2024 | Prospective | NBI endoscopy | EB-03 prototype | 167 UC pts | 8853 images | Diagnosing vascular healing and predicting outcomes in UC |
| Three endoscopists (expertise N/R but registered) |
Ogata et al. [ | 2024 | Prospective | WLE | EB-UC2 prototype | 110 UC pts in clinical remission | 11,472 images | Predicting clinical relapse during 12-month follow-up |
| Two expert endoscopists and six non-specialist endoscopists |
RHI: Robarts Histopathology Index; NHI: Nancy Histological Index; PHRI: PICaSSO Histologic Remission Index; EC: endocytoscopy; WLE: white-light endoscopy; CE: virtual chromoendoscopy; N/R: not reported; UC: ulcerative colitis; AUROC: area under the receiver operating characteristic; IBD: inflammatory bowel disease; CD: Crohn’s disease; CAD: computer-aided detection; MES: Mayo endoscopic score; UCEIS: Ulcerative Colitis Endoscopic Index of Severity; GS: Geboes score; HR: hazard ratio; SWE: single-wavelength endoscopy; and NBI: narrow-band imaging.
References
1. Torres, J.; Mehandru, S.; Colombel, J.F.; Peyrin-Biroulet, L. Crohn’s disease. Lancet; 2017; 389, pp. 1741-1755. [DOI: https://dx.doi.org/10.1016/S0140-6736(16)31711-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27914655]
2. Ungaro, R.; Mehandru, S.; Allen, P.B.; Peyrin-Biroulet, L.; Colombel, J.F. Ulcerative colitis. Lancet; 2017; 389, pp. 1756-1770. [DOI: https://dx.doi.org/10.1016/S0140-6736(16)32126-2]
3. Gros, B.K.G.; Kaplan, G.G. Ulcerative colitis in adults: A review. JAMA; 2023; 330, pp. 951-965. [DOI: https://dx.doi.org/10.1001/jama.2023.15389] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37698559]
4. Neurath, M.F.; Vieth, M. Different levels of healing in inflammatory bowel diseases: Mucosal, histologial, transmural, barrier and complete healing. Gut; 2023; 72, pp. 2164-2183. [DOI: https://dx.doi.org/10.1136/gutjnl-2023-329964] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37640443]
5. Yoon, H.; Jangi, S.; Dulai, P.S.; Boland, B.S.; Prokop, L.J.; Jairath, V.; Feagan, B.G.; Sandborn, W.J.; Singh, S. Incremental Benefit of Achieving Endoscopic and Histologic Remission in Patients With Ulcerative Colitis: A Systematic Review and Meta-Analysis. Gastroenterology; 2020; 159, pp. 1262-1275.e7. [DOI: https://dx.doi.org/10.1053/j.gastro.2020.06.043]
6. Mori, Y.; East, J.E.; Hassan, C.; Halvorsen, N.; Berzin, T.M.; Byrne, M.; von Renteln, D.; Hewett, D.G.; Repici, A.; Ramchandani, M. et al. Benefits and challenges in implementation of artificial intelligence in colonoscopy: World Endoscopy Organization position statement. Dig. Endosc.; 2023; 35, pp. 422-429. [DOI: https://dx.doi.org/10.1111/den.14531]
7. Pal, P.; Pooja, K.; Nabi, Z.; Gupta, R.; Tandan, M.; Rao, G.V.; Reddy, N. Artificial intelligence in endoscopy related to inflammatory bowel disease: A systematic review. Indian. J. Gastroenterol.; 2024; 43, pp. 172-187. [DOI: https://dx.doi.org/10.1007/s12664-024-01531-3]
8. Maeda, Y.; Kudo, S.E.; Kuroki, T.; Iacucci, M. Automated endoscopic diagnosis in IBD The emerging role of artificial intelligence. Gastrointest. Endosc. Clin. N. Am.; 2025; 35, pp. 213-233. [DOI: https://dx.doi.org/10.1016/j.giec.2024.04.012]
9. Santacroce, G.; Zammarchi, I.; Tan, C.K.; Coppola, G.; Varley, R.; Ghosh, S.; Iacucci, M. Present and future of endoscopy precision for inflammatory bowel disease. Dig. Endosc.; 2024; 36, pp. 292-304. [DOI: https://dx.doi.org/10.1111/den.14672]
10. Turner, D.; Ricciuto, A.; Lewis, A.; D’Amico, F.; Dhaliwal, J.; Griffiths, A.M.; Bettenworth, D.; Sandborn, W.J.; Sands, B.E.; Reinisch, W. et al. STRIDE-II: An update on the selecting therapeutic targets in inflammatory bowel disease (STRIDE) initiative of the international organization for the study of IBD (IOIBD): Determining therapeutic goals for treat-to-target strategies in IBD. Gastroenterology; 2021; 160, pp. 1570-1583. [DOI: https://dx.doi.org/10.1053/j.gastro.2020.12.031]
11. Osada, T.; Ohkusa, T.; Yokoyama, T.; Shibuya, T.; Sakamoto, N.; Beppu, K.; Nagahara, A.; Otaka, M.; Ogihara, T.; Watanabe, S. Comparison of several activity indices for the evaluation of endoscopic activity in UC: Inter- and intraobserver consistency. Inflamm. Bowel Dis.; 2010; 16, pp. 192-197. [DOI: https://dx.doi.org/10.1002/ibd.21000]
12. Bisschops, R.; East, J.E.; Hassan, C.; Hazewinkel, Y.; Kamiński, M.F.; Neumann, H.; Pellisé, M.; Antonelli, G.; Bustamante Balen, M.; Coron, E. et al. Advanced imaging for detection and differentiation of colorectal neoplasia: European Society of Gastrointestinal Endoscopy (ESGE) Guideline—Update 2019. Endoscopy; 2019; 51, pp. 1155-1179. Erratum in Endoscopy 2019, 51, C6 [DOI: https://dx.doi.org/10.1055/a-1074-5788]
13. Bryant, R.V.; Burger, D.C.; Delo, J.; Walsh, A.J.; Thomas, S.; von Herbay, A.; Buchel, O.C.; White, L.; Brain, O.; Keshavet, S. et al. Beyond endoscopic mucosal healing in UC: Histological remission better predicts corticosteroid use and hospitalisation over 6 years of follow-up. Gut; 2016; 65, pp. 408-414. [DOI: https://dx.doi.org/10.1136/gutjnl-2015-309598]
14. Simsek, H.D.; Basyigit, S.; Aktas, B.; Simsek, G.G.; Vargol, E.; Kucukazman, M.; Nazlıgul, Y. Assessment of the correlation between endoscopic activity and histological activity in ulcerative colitis patients. Med. Princ. Pract.; 2016; 25, pp. 378-384. [DOI: https://dx.doi.org/10.1159/000445502] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26985865]
15. Kim, D.B.; Lee, K.M.; Lee, J.M.; Chung, Y.Y.; Sung, H.J.; Paik, C.N.; Chung, W.C.; Jung, J.H.; Choi, H.J. Correlation between histological activity and endoscopic, clinical, and serologic activities in patients with ulcerative colitis. Gastroenterol. Res. Pract.; 2016; 2016, 5832051. [DOI: https://dx.doi.org/10.1155/2016/5832051] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26839541]
16. Fluxá, D.; Simian, D.; Flores, L.; Ibáñez, P.; Lubascher, J.; Figueroa, C.; Kronberg, U.; Pizarro, G.; Castro, M.; Piottante, A. et al. Clinical, endoscopic and histological correlation and measures of association in ulcerative colitis. J. Dig. Dis.; 2017; 18, pp. 634-641. [DOI: https://dx.doi.org/10.1111/1751-2980.12546] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28949435]
17. Frieri, G.; Galletti, B.; Di Ruscio, M.; Tittoni, R.; Capannolo, A.; Serva, D.; Latella, G.; Sollima, L.; Leocata, P.; Necozione, S. et al. The prognostic value of histology in ulcerative colitis in clinical remission with mesalazine. Therap. Adv. Gastroenterol.; 2017; 10, pp. 749-759. [DOI: https://dx.doi.org/10.1177/1756283X17722926]
18. Nardone, O.M.; Snir, Y.; Hodson, J.; Cannatelli, R.; Labarile, N.; Siau, K.; Hassan, C.; Yanai, H.; Dotan, I.; Ghosh, S. et al. Advanced technology for assessment of endoscopic and histological activity in ulcerative colitis: A systematic review and meta-analysis. Therap. Adv. Gastroenterol.; 2022; 15, 17562848221092594. [DOI: https://dx.doi.org/10.1177/17562848221092594]
19. Tontini, G.E.; Vecchi, M.; Neurath, M.F. Review article: Newer optical and digital chromoendoscopy techniques vs. dye-based chromoendoscopy for diagnosis and surveillance in inflammatory bowel disease. Aliment. Pharmacol. Ther.; 2013; 38, pp. 1198-1208. [DOI: https://dx.doi.org/10.1111/apt.12508]
20. ASGE Technology Committee Wong Kee Song, L.M.; Adler, D.G.; Chand, B.; Conway, J.D.; Croffie, J.M.B.; Disario, J.A.; Mishkin, D.S.; Shah, R.J.; Somogyi, L. et al. Chromoendoscopy. Gastrointest. Endosc.; 2007; 66, pp. 639-649. [DOI: https://dx.doi.org/10.1016/j.gie.2007.05.029]
21. Dekker, E.; Houwen, B.B.S.L.; Puig, I.; Bustamante-Balén, M.; Coron, E.; Dobru, D.E.; Kuvaev, R.; Neumann, H.; Johnson, G.; Pimentel-Nunes, P. et al. Curriculum for optical diagnosis training in Europe: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement. Endoscopy; 2020; 52, pp. 899-923. Erratum in Endoscopy 2020, 52, C10 [DOI: https://dx.doi.org/10.1055/a-1264-2634] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32882737]
22. Sinonquel, P.; Vermeire, S.; Maes, F.; Bisschops, R. Advanced imaging in gastrointestinal endoscopy: A literature review of the current state of the art. GE Port. J. Gastroenterol.; 2022; 30, pp. 175-191. [DOI: https://dx.doi.org/10.1159/000527083] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37387720]
23. Gui, X.; Bazarova, A.; Del Amor, R.; de Hertogh, G.; Villanacci, V.; Zardo, D.; Parigi, T.L.; Røyset, E.S.; Shivaji, U.N.; Monica, M.A.T. et al. PICaSSO Histologic Remission Index (PHRI) in ulcerative colitis: Development of a novel simplified histological score for monitoring mucosal healing and predicting clinical outcomes and its applicability in an artificial intelligence system. Gut; 2022; 71, pp. 889-898. [DOI: https://dx.doi.org/10.1136/gutjnl-2021-326376] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35173041]
24. Iacucci, M.; Smith, S.C.L.; Bazarova, A.; Shivaji, U.N.; Bhandari, P.; Cannatelli, R.; Daperno, M.; Ferraz, J.; Goetz, M.; Gui, X. et al. An International Multicenter real-life prospective study of electronic chromoendoscopy score PICaSSO in Ulcerative Colitis. Gastroenterology; 2021; 160, pp. 1558-1569.e8. [DOI: https://dx.doi.org/10.1053/j.gastro.2020.12.024]
25. Cannatelli, R.; Bazarova, A.; Furfaro, F.; Parigi, T.L.; Zardo, D.; Nardone, O.M.; Spaggiari, P.; Villanacci, V.; Cadei, M.; Labarile, N. et al. Reproducibility of the electronic chromoendoscopy PICaSSO score (Paddington International Virtual ChromoendoScopy ScOre) in ulcerative colitis using multiple endoscopic platforms: A prospective multicenter international study (with video). Gastrointest. Endosc.; 2022; 96, pp. 73-83. [DOI: https://dx.doi.org/10.1016/j.gie.2022.02.012]
26. Tontini, G.E.; Mudter, J.; Vieth, M.; Atreya, R.; Günther, C.; Zopf, Y.; Wildner, D.; Kiesslich, R.; Vecchi, M.; Neurath, M.F. et al. Confocal laser endomicroscopy for the differential diagnosis of ulcerative colitis and Crohn’s disease: A pilot study. Endoscopy; 2015; 47, pp. 437-443. [DOI: https://dx.doi.org/10.1055/s-0034-1391226]
27. Buchner, A.M. Confocal laser endomicroscopy in the evaluation of inflammatory Bowel Disease. Inflamm. Bowel Dis.; 2019; 25, pp. 1302-1312. [DOI: https://dx.doi.org/10.1093/ibd/izz021]
28. Li, C.Q.; Xie, X.J.; Yu, T.; Gu, X.M.; Zuo, X.L.; Zhou, C.J.; Huang, W.Q.; Chen, H.; Li, Y.Q. Classification of inflammation activity in ulcerative colitis by confocal laser endomicroscopy. Am. J. Gastroenterol.; 2010; 105, pp. 1391-1396. [DOI: https://dx.doi.org/10.1038/ajg.2009.664]
29. Neumann, H.; Vieth, M.; Atreya, R.; Grauer, M.; Siebler, J.; Bernatik, T.; Neurath, M.F.; Mudter, J. Assessment of Crohn’s disease activity by confocal laser endomicroscopy. Inflamm. Bowel Dis.; 2012; 18, pp. 2261-2269. [DOI: https://dx.doi.org/10.1002/ibd.22907]
30. Karstensen, J.G.; Saftoiu, A.; Brynskov, J.; Hendel, J.; Ciocalteu, A.; Klausen, P.; Klausen, T.W.; Riis, L.B.; Vilmann, P. Confocal laser endomicroscopy in ulcerative colitis: A longitudinal study of endomicroscopic changes and response to medical therapy (with videos). Gastrointest. Endosc.; 2016; 84, pp. 279-286.e1. [DOI: https://dx.doi.org/10.1016/j.gie.2016.01.069]
31. Buda, A.; Hatem, G.; Neumann, H.; D’Incà, R.; Mescoli, C.; Piselli, P.; Jackson, J.; Bruno, M.; Sturniolo, G.C. Confocal laser endomicroscopy for prediction of disease relapse in ulcerative colitis: A pilot study. J. Crohns Colitis; 2014; 8, pp. 304-311. [DOI: https://dx.doi.org/10.1016/j.crohns.2013.09.005] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24094597]
32. Neumann, H.; Kudo, S.E.; Kiesslich, R.; Neurath, M.F. Advanced colonoscopic imaging using endocytoscopy. Dig. Endosc.; 2015; 27, pp. 232-238. [DOI: https://dx.doi.org/10.1111/den.12395] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25311804]
33. Bessho, R.; Kanai, T.; Hosoe, N.; Kobayashi, T.; Takayama, T.; Inoue, N.; Mukai, M.; Ogata, H.; Hibi, T. Correlation between endocytoscopy and conventional histopathology in microstructural features of ulcerative colitis. J. Gastroenterol.; 2011; 46, pp. 1197-1202. [DOI: https://dx.doi.org/10.1007/s00535-011-0439-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21805068]
34. Vitali, F.; Morgenstern, N.; Eckstein, M.; Atreya, R.; Waldner, M.; Hartmann, A.; Neurath, M.F.; Rath, T. Endocytoscopy for assessing histologic inflammation in ulcerative colitis: Development and prospective validation of the ELECT (ErLangen Endocytoscopy in ColiTis) score (with videos). Gastrointest. Endosc.; 2023; 97, pp. 100-111.e1. [DOI: https://dx.doi.org/10.1016/j.gie.2022.08.023]
35. Kiesslich, R.; Duckworth, C.A.; Moussata, D.; Gloeckner, A.; Lim, L.G.; Goetz, M.; Pritchard, D.M.; Galle, P.R.; Neurath, M.F.; Watson, A.J.M. Local barrier dysfunction identified by confocal laser endomicroscopy predicts relapse in inflammatory bowel disease. Gut; 2012; 61, pp. 1146-1153. [DOI: https://dx.doi.org/10.1136/gutjnl-2011-300695]
36. Neumann, H.; Vieth, M.; Neurath, M.F.; Atreya, R. Endocytoscopy allows accurate in vivo differentiation of mucosal inflammatory cells in IBD: A pilot study. Inflamm. Bowel Dis.; 2013; 19, pp. 356-362. [DOI: https://dx.doi.org/10.1002/ibd.23025]
37. Atreya, R.; Neumann, H.; Neufert, C.; Waldner, M.J.; Billmeier, U.; Zopf, Y.; Willma, M.; App, C.; Münster, T.; Kessler, H. et al. In vivo imaging using fluorescent antibodies to tumor necrosis factor predicts therapeutic response in Crohn’s disease. Nat. Med.; 2014; 20, pp. 313-318. [DOI: https://dx.doi.org/10.1038/nm.3462]
38. Rath, T.; Bojarski, C.; Neurath, M.F.; Atreya, R. Molecular imaging of mucosal a4b7 integrin expression with the fluorescent anti-adhesion antibody vedolizumab in Crohn’s disease. Gastrointest. Endosc.; 2017; 86, pp. 406-408. [DOI: https://dx.doi.org/10.1016/j.gie.2017.01.012]
39. Ahmad, H.A.; East, J.E.; Panaccione, R.; Travis, S.; Canavan, J.B.; Usiskin, K.; Byrne, M.F. Artificial intelligence in inflammatory bowel disease endoscopy: Implications for clinical trials. J. Crohns Colitis; 2023; 17, pp. 1342-1353. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjad029]
40. Oh, D.J.; Hwang, Y.; Kim, S.H.; Nam, J.H.; Jung, M.K.; Lim, Y.J. Reading of small bowel capsule endoscopy after frame reduction using an artificial intelligence algorithm. BMC Gastroenterol.; 2024; 24, 80. [DOI: https://dx.doi.org/10.1186/s12876-024-03156-4]
41. Sutton, T.; Zai Ane, O.R.; Goebel, R.; Baumgart, D.C. Artificial intelligence enabled automated diagnosis and grading of ulcerative colitis endoscopy images. Sci. Rep.; 2022; 12, 2748. [DOI: https://dx.doi.org/10.1038/s41598-022-06726-2]
42. Sharma, A.; Kumar, R.; Garg, P. Deep learning-based prediction model for diagnosing gastrointestinal diseases using endoscopy images. Int. J. Med. Inform.; 2023; 177, 105142. [DOI: https://dx.doi.org/10.1016/j.ijmedinf.2023.105142] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37422969]
43. Guimarães, P.; Finkler, H.; Reichert, M.C.; Zimmer, V.; Grünhage, F.; Krawczyk, M.; Lammert, F.; Keller, A.; Casper, M. Artificial-intelligence-based decision support tools for the differential diagnosis of colitis. Eur. J. Clin. Investig.; 2023; 53, e13960. [DOI: https://dx.doi.org/10.1111/eci.13960]
44. Kim, J.M.; Kang, J.G.; Kim, S.; Cheon, J.H. Deep-learning system for real-time differentiation between Crohn’s disease, intestinal Behçet’s disease, and intestinal tuberculosis. J. Gastroenterol. Hepatol.; 2021; 36, pp. 2141-2148. [DOI: https://dx.doi.org/10.1111/jgh.15433]
45. Tong, Y.; Lu, K.; Yang, Y.; Li, J.; Lin, Y.; Wu, D.; Yang, A.; Li, Y.; Yu, S.; Qian, J. Can natural language processing helpdifferentiate inflammatory intestinal diseases in China? Models applying random forest and convolutional neural network approaches. BMC Med. Inform. Decis. Mak.; 2020; 20, 248. [DOI: https://dx.doi.org/10.1186/s12911-020-01277-w] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32993636]
46. Park, J.J.; Park, S.J.; Hong, S.P.; Kim, T.I.; Kim, W.H.; Cheon, J.H. Su1932 Differential diagnosis between intestinal tuberculosis and Crohn’s disease by ileocolonoscopic findings. Gastroenterology; 2012; 142, S-539. [DOI: https://dx.doi.org/10.1016/S0016-5085(12)62071-X]
47. Lu, K.; Tong, Y.; Yu, S.; Lin, Y.; Yang, Y.; Xu, H.; Li, Y.; Yu, S. Building a trustworthy AI differential diagnosis application for Crohn’s disease and intestinal tuberculosis. BMC Med. Inform. Decis. Mak.; 2023; 23, 160. [DOI: https://dx.doi.org/10.1186/s12911-023-02257-6] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37582768]
48. Lu, Y.; Chen, Y.; Peng, X.; Yao, J.; Zhong, W.; Li, C.; Zhi, M. Development and validation of a new algorithm model for differential diagnosis between Crohn’s disease and intestinal tuberculosis: A combination of laboratory, imaging and endoscopic characteristics. BMC Gastroenterol.; 2021; 21, 291. [DOI: https://dx.doi.org/10.1186/s12876-021-01838-x]
49. Ruan, G.; Qi, J.; Cheng, Y.; Liu, R.; Zhang, B.; Zhi, M.; Chen, J.; Xiao, F.; Shen, X.; Fan, L. et al. Development and validation of a deep neural network for accurate identification of endoscopic images from patients with ulcerative colitis and Crohn’s disease. Front. Med.; 2022; 9, 854677. [DOI: https://dx.doi.org/10.3389/fmed.2022.854677]
50. Wang, L.; Chen, L.; Wang, X.; Liu, K.; Li, T.; Yu, Y.; Han, J.; Xing, S.; Xu, J.; Tian, D. et al. Development of a Convolutional Neural Network-Based Colonoscopy Image Assessment Model for Differentiating Crohn’s Disease and Ulcerative Colitis. Front. Med.; 2022; 9, 789862. [DOI: https://dx.doi.org/10.3389/fmed.2022.789862]
51. Chierici, M.; Puica, N.; Pozzi, M.; Capistrano, A.; Donzella, M.D.; Colangelo, A.; Osmani, V.; Jurman, G. Automatically detecting Crohn’s disease and ulcerative colitis from endoscopic imaging. BMC Med. Inform. Decis. Mak.; 2022; 22, (Suppl. S6), 300. [DOI: https://dx.doi.org/10.1186/s12911-022-02043-w] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36401328]
52. Quénéhervé, L.; David, G.; Bourreille, A.; Hardouin, J.B.; Rahmi, G.; Neunlist, M.; Brégeon, J.; Coron, E. Quantitative assessment of mucosal architecture using computer based analysis of confocal laser endomicroscopy in inflammatory bowel diseases. Gastrointest. Endosc.; 2019; 89, pp. 626-636. [DOI: https://dx.doi.org/10.1016/j.gie.2018.08.006]
53. Higuchi, N.; Hiraga, H.; Sasaki, Y.; Hiraga, N.; Igarashi, S.; Hasui, K.; Ogasawara, K.; Maeda, T.; Murai, Y.; Tatsuta, T. et al. Automated evaluation of colon capsule endoscopic severity of ulcerative colitis using ResNet50. PLoS ONE; 2022; 17, e0269728. [DOI: https://dx.doi.org/10.1371/journal.pone.0269728]
54. Majtner, T.; Brodersen, J.B.; Herp, J.; Kjeldsen, J.; Halling, M.L.; Jensen, M.D. A deep learning framework for autonomous detection and classification of Crohn’s disease lesions in the small bowel and colon with capsule endoscopy. Endosc. Intern. Open; 2021; 9, pp. E1361-E1370. [DOI: https://dx.doi.org/10.1055/a-1507-4980]
55. Brodersen, J.B.; Jensen, M.D.; Leenhardt, R.; Kjeldsen, J.; Histace, A.; Knudsen, T.; Dray, X. Artificial intelligence-assisted analysis of pan-enteric capsule endoscopy in patients with suspected Crohn’s disease: A study on diagnostic performance. J. Crohns Colitis; 2024; 18, pp. 75-81. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjad131] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37527554]
56. Charisis, V.S.; Hadjileontiadis, L.J. Potential of hybrid adaptive filtering in inflammatory lesion detection from capsule endoscopy images. World J. Gastroenterol.; 2016; 22, pp. 8641-8657. [DOI: https://dx.doi.org/10.3748/wjg.v22.i39.8641]
57. Aoki, T.; Yamada, A.; Aoyama, K.; Saito, H.; Tsuboi, A.; Nakada, A.; Niikura, R.; Fujishiro, M.; Oka, S.; Ishihara, S. et al. Automatic detection of erosions and ulcerations in wireless capsule endoscopy images based on a deep convolutional neural network. Gastrointest. Endosc.; 2019; 89, pp. 357-363.e2. [DOI: https://dx.doi.org/10.1016/j.gie.2018.10.027]
58. Klang, E.; Barash, Y.; Margalit, R.Y.; Soffer, S.; Shimon, O.; Albshesh, A.; Ben-Horin, S.; Amitai, M.M.; Eliakim, R.; Kopylov, U. Deep learning algorithms for automated detection of Crohn’s disease ulcers by video capsule endoscopy. Gastrointest. Endosc.; 2020; 91, pp. 606-612.e2. [DOI: https://dx.doi.org/10.1016/j.gie.2019.11.012]
59. Barash, Y.; Azaria, L.; Soffer, S.; Yehuda, R.M.; Shlomi, O.; Ben-Horin, S.; Eliakim, R.; Klang, E.; Kopylov, U. Ulcer severity grading in video capsule images of patients with Crohn’s disease: An ordinal neural network solution. Gastrointest. Endosc.; 2021; 93, pp. 187-192. [DOI: https://dx.doi.org/10.1016/j.gie.2020.05.066]
60. Klang, E.; Grinman, A.; Soffer, A.; Yehuda, R.M.; Barzilay, O.; Amitai, M.M.; Konen, E.; Ben-Horin, S.; Eliakim, R.; Barash, Y. et al. Automated detection of Crohn’s disease intestinal strictures on capsule endoscopy images using deep neural networks. J. Crohn Colitis.; 2021; 15, pp. 749-756. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjaa234]
61. De Maissin, A.; Vallee, R.; Flamant, M.; Fondain-Bossiere, M.; Le Berre, C.; Coutrot, A.; Normand, N.; Mouchère, H.; Coudol, S.; Trang, C. et al. Multi-expert annotation of Crohn’s disease images of the small bowel for automatic detection using a convolutional recurrent attention neural network. Endosc. Int. Open; 2021; 9, pp. E1136-E1144. [DOI: https://dx.doi.org/10.1055/a-1468-3964] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34222640]
62. Ferreira, J.P.S.; de Mascarenhas Saraiva, M.J.; Afonso, J.P.L.; Carneiro Ribeiro, T.F.; Casal Cardoso, H.M.; Ribeiro Andrade, A.P.; de Mascarenhas Saraiva, M.N.G.; Lages Parente, M.P.; Jorge, R.N.; Oliveira Lopes, S.I. et al. Identification of ulcers and erosions by the novel PillCam Crohn’s capsule using a convolutional neural network: A multicentre pilot study. J. Crohns Colitis; 2022; 16, pp. 169-172. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjab117] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34228113]
63. Kratter, T.; Shapira, N.; Lev, Y.; Mauda, O.; Moshkovitz, Y.; Shitrit, R.; Konyo, S.; Ukashi, O.; Dar, L.; Shlomi, O. et al. Deep learning multi-domain model provides accurate detection and grading of mucosal ulcers in different capsule endoscopy types. Diagnostics; 2022; 12, 2490. [DOI: https://dx.doi.org/10.3390/diagnostics12102490]
64. Ribeiro, T.; Mascarenhas, M.; Afonso, J.; Cardoso, H.; Andrade, P.; Lopes, S.; Ferreira, J.; Mascarenhas Saraiva, M.; Macedo, G. Artificial intelligence and colon capsule endoscopy: Automatic detection of ulcers and erosions using a convolutional neural network. J. Gastroenterol. Hepatol.; 2022; 37, pp. 2282-2288. [DOI: https://dx.doi.org/10.1111/jgh.16011] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36181257]
65. Wang, S.; Xing, Y.; Zhang, L.; Gao, H.; Zhang, H. A systematic evaluation and optimization of automatic detection of ulcers in wireless capsule endoscopy on a large dataset using deep convolutional neural networks. Phys. Med. Biol.; 2019; 64, 235014. [DOI: https://dx.doi.org/10.1088/1361-6560/ab5086]
66. Christensen, B.; Rubin, D.T. Understanding endoscopic disease activity in IBD: How to incorporate it into practice. Curr. Gastroenterol. Rep.; 2016; 18, 5. [DOI: https://dx.doi.org/10.1007/s11894-015-0477-6]
67. Ikeya, K.; Hanai, H.; Sugimoto, K.; Osawa, S.; Kawasaki, S.; Iida, T.; Maruyama, Y.; Watanabe, F. The ulcerative colitis endoscopic index of severity more accurately reflects clinical outcomes and long-term prognosis than the Mayo endoscopic score. J. Crohns Colitis; 2016; 10, pp. 286-295. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjv210]
68. Kim, J.E.; Choi, Y.H.; Lee, Y.C.; Seong, G.; Song, J.H.; Kim, T.J.; Kim, E.R.; Hong, S.N.; Chang, D.K.; Kim, Y.H. et al. Deep learning model for distinguishing Mayo endoscopic subscore 0 and 1 in patients with ulcerative colitis. Sci. Rep.; 2023; 13, 11351. [DOI: https://dx.doi.org/10.1038/s41598-023-38206-6]
69. Wang, G.; Zahng, S.; Li, J.; Zhao, K.; Ding, Q.; Tian, D.; Li, R.; Zou, F.; Yu, Q. CB-HRNet: A Class-Balanced High-Resolution Network for the evaluation of endoscopic activity in patients with ulcerative colitis. Clin. Transl. Sci.; 2023; 16, pp. 1421-1430. [DOI: https://dx.doi.org/10.1111/cts.13542]
70. Polat, G.; Kani, H.T.; Ergenc, I.; Alahdab, Y.O.; Temizel, A.; Atug, O. Improving the Computer-Aided Estimation of Ulcerative Colitis Severity According to Mayo Endoscopic Score by Using Regression-Based Deep Learning. Inflamm. Bowel Dis.; 2023; 29, pp. 1431-1439. [DOI: https://dx.doi.org/10.1093/ibd/izac226]
71. Qi, J.; Ruan, G.; Ping, Y.; Xiao, Z.; Liu, K.; Cheng, Y.; Liu, R.; Zhang, B.; Zhi, M.; Chen, J. et al. Development and validation of a deep learning-based approach to predict the Mayo endoscopic score of ulcerative colitis. Ther. Adv. Gastroenterol.; 2023; 16, 17562848231170945. [DOI: https://dx.doi.org/10.1177/17562848231170945]
72. Turan, M.; Durmus, F. UC-NfNet: Deep learning-enabled assessment of ulcerative colitis from colonoscopy images. Med. Image Anal.; 2022; 82, 102587. [DOI: https://dx.doi.org/10.1016/j.media.2022.102587] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36058054]
73. Iacucci, M.; Cannatelli, R.; Parigi, T.L.; Nardone, O.M.; Tontini, G.E.; Labarile, N.; Buda, A.; Rimondi, A.; Bazarova, A.; Bisschops, R. et al. A virtual chromoendoscopy artificial intelligence system to detect endoscopic and histologic activity/remission and predict clinical outcomes in ulcerative colitis. Endoscopy; 2023; 55, pp. 332-341. [DOI: https://dx.doi.org/10.1055/a-1960-3645]
74. Patel, M.; Gulati, S.; Iqbal, F.; Hayee, B. Rapid development of accurate artificial intelligence scoring for colitis disease activity using applied data science techniques. Endosc. Int. Open; 2022; 10, pp. E539-E543. [DOI: https://dx.doi.org/10.1055/a-1790-6201] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35433223]
75. Takabayashi, K.; Kobayashi, T.; Matsuoka, K.; Levesque, B.G.; Kawamura, T.; Tanaka, K.; Kadota, T.; Bise, R.; Uchida, S.; Kanai, T. et al. Artificial intelligence quantifying endoscopic severity of ulcerative colitis in gradation scale. Dig. Endosc.; 2024; 36, pp. 582-590. [DOI: https://dx.doi.org/10.1111/den.14677] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37690125]
76. Lo, B.; Liu, Z.; Bendtsen, F.; Igel, C.; Vind, I.; Burisch, J. High Accuracy in Classifying Endoscopic Severity in Ulcerative Colitis Using Convolutional Neural Network. Am. J. Gastroenterol.; 2022; 117, pp. 1648-1654. [DOI: https://dx.doi.org/10.14309/ajg.0000000000001904]
77. Yao, H.; Najarian, K.; Gryak, J.; Bishu, S.; Rice, M.D.; Waljee, A.K.; Wilkins, H.J.; Stidham, R.W. Fully automated endoscopic disease activity assessment in ulcerative colitis. Gastrointest. Endosc.; 2021; 93, pp. 728-736.e1. [DOI: https://dx.doi.org/10.1016/j.gie.2020.08.011]
78. Stidham, R.W.; Liu, W.; Bishu, S.; Rice, M.D.; Higgins, P.D.R.; Zhu, J.; Nallamothu, B.K.; Waljee, A.K. Performance of a Deep Learning Model vs Human Reviewers in Grading Endoscopic Disease Severity of Patients With Ulcerative Colitis. JAMA Netw. Open; 2019; 2, e193963. [DOI: https://dx.doi.org/10.1001/jamanetworkopen.2019.3963]
79. Byrne, M.F.; Panaccione, R.; East, J.E.; Iacucci, M.; Parsa, N.; Kalapala, R.; Reddy, D.N.; Rughwani, H.R.; Singh, A.P.; Berry, S.K. et al. Application of Deep Learning Models to Improve Ulcerative Colitis Endoscopic Disease Activity Scoring Under Multiple Scoring Systems. J. Crohns Colitis; 2023; 17, pp. 463-471. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjac152]
80. Ozawa, T.; Ishihara, S.; Fujishiro, M.; Saito, H.; Kumagai, Y.; Shichijo, S.; Aoyama, K.; Tada, T. Novel computer-assisted diagnosis system for endoscopic disease activity in patients with ulcerative colitis. Gastrointest. Endosc.; 2019; 89, pp. 416-421.e1. [DOI: https://dx.doi.org/10.1016/j.gie.2018.10.020]
81. Huang, T.Y.; Zhan, S.Q.; Chen, P.J.; Yang, C.W.; Lu, H.H.S. Accurate diagnosis of endoscopic mucosal healing in ulcerative colitis using deep learning and machine learning. J. Chin. Med. Assoc.; 2021; 84, pp. 678-681. [DOI: https://dx.doi.org/10.1097/JCMA.0000000000000559]
82. Barreiro-de Acosta, M.; Vallejo, N.; de la Iglesia, D.; Uribarri, L.; Bastón, I.; Ferreiro-Iglesias, R.; Lorenzo, A.; Domínguez-Muñoz, J.E. Evaluation of the risk of relapse in ulcerative colitis according to the degree of mucosal healing (Mayo 0 vs 1): A longitudinal cohort study. J. Crohns Colitis; 2016; 10, pp. 13-19. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjv158]
83. Bhambhvani, H.P.; Zamora, A. Deep learning enabled classification of Mayo endoscopic subscore in patients with ulcerative colitis. Eur. J. Gastroenterol. Hepatol.; 2021; 33, pp. 645-649. [DOI: https://dx.doi.org/10.1097/MEG.0000000000001952] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33079775]
84. Gutierrez Becker, B.; Arcadu, F.; Thalhammer, A.; Gamez Serna, C.; Feehan, O.; Drawnel, F.; Oh, Y.S.; Prunotto, M. Training and deploying a deep learning model for endoscopic severity grading in ulcerative colitis using multicenter clinical trial data. Ther. Adv. Gastrointest. Endosc.; 2021; 14, 2631774521990623. [DOI: https://dx.doi.org/10.1177/2631774521990623] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33718871]
85. Gottlieb, K.; Requa, J.; Karnes, W.; Gudivada, R.C.; Shen, J.; Rael, E.; Arora, V.; Dao, T.; Ninh, A.; McGill, J. Central Reading of Ulcerative Colitis Clinical Trial Videos Using Neural Networks. Gastroenterology; 2021; 160, pp. 710-719.e2. [DOI: https://dx.doi.org/10.1053/j.gastro.2020.10.024]
86. Fan, Y.; Mu, R.; Xu, H.; Xie, C.; Zhang, Y.; Liu, L.; Wang, L.; Shi, H.; Hu, Y.; Ren, J. et al. Novel deep learning-based computer-aided diagnosis system for predicting inflammatory activity in ulcerative colitis. Gastrointest. Endosc.; 2023; 97, pp. 335-346. [DOI: https://dx.doi.org/10.1016/j.gie.2022.08.015] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35985375]
87. Stidham, R.W.; Cai, L.; Cheng, S.; Rajaei, F.; Hiatt, T.; Wittrup, E.; Rice, M.D.; Bishu, S.; Wehkamp, J.; Schultz, W. et al. Using computer vision to improve endoscopic disease quantification in therapeutic clinical trials of ulcerative colitis. Gastroenterology; 2024; 166, pp. 155-167.e2. [DOI: https://dx.doi.org/10.1053/j.gastro.2023.09.049]
88. Gutierrez Becker, B.; Fraessle, S.; Yao, H.; Luscher, J.; Girycki, R.; Machura, B.; Czornik, J.; Goslinsky, J.; Pitura, M.; Levitte, S. et al. Ulcerative Colitis Severity Classification and Localised Extent (UCSCALE): An Artificial Intelligence scoring system for a spatial assessment of disease severity in Ulcerative Colitis. J. Crohns Colitis; 2025; 19, jjae187. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjae187]
89. Akiyama, S.; Sakamoto, T.; Kobayashi, M.; Matsubara, D.; Tsuchiya, K. Clinical usefulness of hypoxia imaging colonoscopy for the objective measurement of ulcerative colitis disease activity. Gastrointest. Endosc.; 2024; 99, pp. 1006-1016.e4. [DOI: https://dx.doi.org/10.1016/j.gie.2023.12.035]
90. Martins, M.; Mascarenhas, M.; Afonso, J.; Ribeiro, T.; Cardoso, P.; Mendes, F.; Cardoso, H.; Andrade, P.; Ferreira, J.; Macedo, G. Deep-Learning and Device-Assisted Enteroscopy: Automatic Panendoscopic Detection of Ulcers and Erosions. Medicina; 2023; 59, 172. [DOI: https://dx.doi.org/10.3390/medicina59010172]
91. Xie, W.; Ju, J.; Liang, P.; Mei, Q.; Wang, A.; Liu, Q.; Liu, X.; Wu, J.; Yang, X.; Zhu, N. et al. Deep learning-based lesion detection and severity grading of small-bowel Crohn’s disease ulcers on double-balloon endoscopy images. Gastrointest. Endosc.; 2024; 99, pp. 767-777.e5. [DOI: https://dx.doi.org/10.1016/j.gie.2023.11.059] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38065509]
92. Udristoiu, A.L.; Stefanescu, D.; Gruionu, G.; Iacob, A.V.; Karstensen, J.G.; Vilman, P.; Saftoiu, A. Deep Learning Algorithm for the Confirmation of Mucosal Healing in Crohn’s Disease, Based on Confocal Laser Endomicroscopy Images. J. Gastrointest. Liver Dis.; 2021; 30, pp. 59-65. [DOI: https://dx.doi.org/10.15403/jgld-3212]
93. Colombel, J.F.; Keir, M.E.; Scherl, A.; Zhao, R.; de Hertogh, G.; Faubion, W.A.; Lu, T.T. Discrepancies between patient-reported outcomes, and endoscopic and histological appearance in UC. Gut; 2016; 66, pp. 2063-2068. [DOI: https://dx.doi.org/10.1136/gutjnl-2016-312307] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27590995]
94. Rutter, M.; Saunders, B.; Wilkinson, K.; Rumbles, S.; Schofield, G.; Kamm, M.; Williams, C.; Price, A.; Talbot, I.; Forbes, A. Severity of inflammation is a risk factor for colorectal neoplasia in ulcerative colitis. Gastroenterology; 2004; 126, pp. 451-459. [DOI: https://dx.doi.org/10.1053/j.gastro.2003.11.010] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/14762782]
95. Iacucci, M.; Fort Gasia, M.; Hassan, C.; Panaccione, R.; Kaplan, G.G.; Ghosh, S.; Gui, X. Complete mucosal healing defined by endoscopic Mayo subscore still demonstrates abnormalities by novel high definition colonoscopy and refined histological gradings. Endoscopy; 2015; 47, pp. 726-734. [DOI: https://dx.doi.org/10.1055/s-0034-1391863]
96. Pessarelli, T.; Tontini, G.E.; Neumann, H. Advanced Endoscopic Imaging for Assessing Mucosal Healing and Histologic Remission in Inflammatory Bowel Diseases. Gastrointest. Endosc. Clin. N. Am.; 2025; 35, pp. 159-177. [DOI: https://dx.doi.org/10.1016/j.giec.2024.04.008]
97. Kiesslich, R.; Neurath, M.F. Advanced endoscopy imaging in inflammatory bowel diseases. Gastrointest. Endosc.; 2017; 85, pp. 496-508. [DOI: https://dx.doi.org/10.1016/j.gie.2016.10.034]
98. Iacucci, M.; Daperno, M.; Lazarev, M.; Arsenascu, R.; Tontini, G.E.; Akinola, O.; Gui, X.S.; Villanacci, V.; Goetz, M.; Lowerison, M. et al. Development and reliability of the new endoscopic virtual chromoendoscopy score: The PICaSSO (Paddington International Virtual ChromoendoScopy ScOre) in ulcerative colitis. Gastrointest. Endosc.; 2017; 86, pp. 1118-1127. [DOI: https://dx.doi.org/10.1016/j.gie.2017.03.012]
99. Nakazato, Y.; Naganuma, M.; Sugimoto, S.; Bessho, R.; Arai, M.; Kiyohara, H.; Ono, K.; Nanki, K.; Mutaguchi, M.; Mizuno, S. et al. Endocytoscopy can be used to assess histological healing in ulcerative colitis. Endoscopy; 2017; 49, pp. 560-563. [DOI: https://dx.doi.org/10.1055/s-0043-106556]
100. Maeda, Y.; Kudo, S.E.; Mori, Y.; Misawa, M.; Ogata, N.; Sasanuma, S.; Wakamura, K.; Oda, M.; Mori, K.; Ohtsuka, K. Fully automated diagnostic system with artificial intelligence using endocytoscopy to identify the presence of histologic inflammation associated with ulcerative colitis (with video). Gastrointest. Endosc.; 2019; 89, pp. 408-415. [DOI: https://dx.doi.org/10.1016/j.gie.2018.09.024]
101. Omori, T.; Yamamoto, T.; Murasugi, S.; Koroku, M.; Yonezawa, M.; Nonaka, K.; Nagashima, Y.; Nakamura, S.; Tokushige, K. Comparison of Endoscopic and Artificial Intelligence Diagnoses for Predicting the Histological Healing of Ulcerative Colitis in a Real-World Clinical Setting. Crohns Colitis 360; 2024; 6, otae005. [DOI: https://dx.doi.org/10.1093/crocol/otae005] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38419859]
102. Takenaka, K.; Ohtsuka, K.; Fujii, T.; Negi, M.; Suzuki, K.; Shimizu, H.; Oshima, S.; Akiyama, S.; Motobayashi, M.; Nagahori, M. et al. Development and validation of a deep neural network for accurate evaluation of endoscopic images from patients with ulcerative colitis. Gastroenterology; 2020; 158, pp. 2150-2157. [DOI: https://dx.doi.org/10.1053/j.gastro.2020.02.012]
103. Takenaka, K.; Ohtsuka, K.; Fujii, T.; Oshima, S.; Okamoto, R.; Watanabe, M. Deep Neural Network Accurately Predicts Prognosis of Ulcerative Colitis Using Endoscopic Images. Gastroenterology; 2021; 160, pp. 2175-2177.e3. [DOI: https://dx.doi.org/10.1053/j.gastro.2021.01.210] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33485853]
104. Takenaka, K.; Fujii, T.; Kawamoto, A.; Suzuki, K.; Shimizu, H.; Maeyashiki, C.; Yamaji, O.; Motobayashi, M.; Igarashi, A.; Hanazawa, R. et al. Deep neural network for video colonoscopy of ulcerative colitis: A cross-sectional study. Lancet Gastroenterol. Hepatol.; 2022; 7, pp. 230-237. [DOI: https://dx.doi.org/10.1016/S2468-1253(21)00372-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34856196]
105. Bossuyt, P.; Nakase, H.; Vermeire, S.; de Hertogh, G.; Eelbode, T.; Ferrante, M.; Hasegawa, T.; Willekens, H.; Ikemoto, Y.; Makino, T. et al. Automatic, computer-aided determination of endoscopic and histological inflammation in patients with mild to moderate ulcerative colitis based on red density. Gut; 2020; 69, pp. 1778-1786. [DOI: https://dx.doi.org/10.1136/gutjnl-2019-320056]
106. Sinonquel, P.; Bossuyt, P.; Sabino, J.P.G.; Verstockt, B.; Makino, T.; Vermeire, S.; Bisschops, R. Long-term follow-up of the red density pilot trial: A basis for long-term prediction of sustained clinical remission in ulcerative colitis?. Endosc. Int. Open; 2023; 11, pp. E880-E884. [DOI: https://dx.doi.org/10.1055/a-2153-7210]
107. Sinonquel, P.; Lenfant, M.; Eelbode, T.; Watanabe, H.; Callaerts, B.; Bossuyt, P.; Verstockt, B.; Sabino, J.P.G.; De Hertogh, G.; Maes, F. et al. Development of an Automated Tool for the Estimation of Histological Remission in Ulcerative Colitis Using Single Wavelength Endoscopy Technology. J. Crohns Colitis; 2024; jjae180.online ahead of print [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjae180]
108. Bossuyt, P.; De Hertogh, G.; Eelbode, T.; Vermeire, S.; Bisschops, R. Computer-Aided Diagnosis With Monochromatic Light Endoscopy for Scoring Histologic Remission in Ulcerative Colitis. Gastroenterology; 2021; 160, pp. 23-25. [DOI: https://dx.doi.org/10.1053/j.gastro.2020.09.053]
109. Maeda, Y.; Kudo, S.E.; Ogata, N.; Misawa, M.; Iacucci, M.; Homma, M.; Nemoto, T.; Takishima, K.; Mochida, K.; Miyachi, H. et al. Evaluation in real-time use of artificial intelligence during colonscopy to predict relapse of ulcerative colitis: A prospective study. Gastrointest. Endosc.; 2022; 95, pp. 747-756.e2. [DOI: https://dx.doi.org/10.1016/j.gie.2021.10.019]
110. Kuroki, T.; Maeda, Y.; Kudo, S.E.; Ogata, N.; Iacucci, M.; Takishima, K.; Ide, Y.; Shibuya, T.; Semba, S.; Kawashima, J. et al. A novel artificial intelligence-assisted “vascular healing” diagnosis for prediction of future clinical relapse in patients with ulcerative colitis: A prospective cohort study (with video). Gastrointest. Endosc.; 2024; 100, pp. 97-108. [DOI: https://dx.doi.org/10.1016/j.gie.2024.01.010]
111. Ogata, N.; Maeda, Y.; Misawa, M.; Takenaka, K.; Takabayashi, K.; Iacucci, M.; Kuroki, T.; Takishima, K.; Sasabe, K.; Niimura, Y. et al. Artificial Intelligence-assisted Video Colonoscopy for Disease Monitoring of Ulcerative Colitis: A Prospective Study. J. Crohns Colitis; 2025; 19, jjae080. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjae080]
112. Eaden, J.A.; Abrams, K.R.; Mayberry, J.F. The risk of colorectal cancer in ulcerative colitis: A meta-analysis. Gut; 2001; 48, pp. 526-535. [DOI: https://dx.doi.org/10.1136/gut.48.4.526] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11247898]
113. Beaugerie, L.; Itzkowitz, S.H. Cancers complicating inflammatory bowel disease. N. Engl. J. Med.; 2015; 372, pp. 1441-1452. [DOI: https://dx.doi.org/10.1056/NEJMra1403718] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25853748]
114. Maeda, Y.; Kudo, S.E.; Ogata, N.; Misawa, M.; Mori, Y.; Mori, K.; Ohtsuka, K. Can artificial intelligence help to detect dysplasia in patients with ulcerative colitis?. Endoscopy; 2021; 53, pp. E273-E274. [DOI: https://dx.doi.org/10.1055/a-1261-2944] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33003217]
115. Fukunaga, S.; Kusaba, Y.; Ohuchi, A.; Nagata, T.; Mitsuyama, K.; Tsuruta, O.; Torimura, T. Is artificial intelligence a superior diagnostician in ulcerative colitis?. Endoscopy; 2021; 53, pp. E75-E76. [DOI: https://dx.doi.org/10.1055/a-1195-1986]
116. Guerrero Vinsard, D.; Fetzer, J.R.; Agrawal, U.; Singh, J.; Damani, D.N.; Sivasubramaniam, P.; Arunachalam, S.P.; Leggett, C.L.; Raffals, L.E.; Coelho-Prabhu, N. Development of an artificial intelligence tool for detecting colorectal lesions in inflammatory bowel disease. iGIE; 2023; 2, pp. 91-101.E6. Available online: https://www.igiejournal.org/article/S2949-7086(23)00042-0/fulltext (accessed on 27 December 2024).
117. Abdelrahim, M.; Siggens, K.; Iwadate, Y.; Maeda, N.; Htet, H.; Bhandari, P. New AI model for neoplasia detection and characterisation in inflammatory bowel disease. Gut; 2024; 73, pp. 725-728. [DOI: https://dx.doi.org/10.1136/gutjnl-2023-330718]
118. Yamamoto, S.; Kinugasa, H.; Hamada, K.; Tomiya, M.; Tanimoto, T.; Ohto, A.; Toda, A.; Takei, D.; Matsubara, M.; Suzuki, S. et al. The diagnostic ability to classify neoplasias occurring in inflammatory bowel disease by artificial intelligence and endoscopists: A pilot study. J. Gastroenterol. Hepatol.; 2022; 37, pp. 1610-1616. [DOI: https://dx.doi.org/10.1111/jgh.15904]
119. Hou, J.K.; Chang, M.; Nguyen, T.; Kramer, J.R.; Richardson, P.; Sansgiry, S.; D’Avolio, L.W.; El-Serag, H.B. Automated identification of surveillance colonoscopy in inflammatory bowel disease using natural language processing. Dig. Dis. Sci.; 2013; 58, pp. 936-941. [DOI: https://dx.doi.org/10.1007/s10620-012-2433-8]
120. Iacucci, M.; Parigi, T.L.; Del Amor, R.; Meseguer, P.; Mandelli, G.; Bozzola, A.; Bazarova, A.; Bhandari, P.; Bisschops, R.; Danese, S. et al. Artificial Intelligence Enabled Histological Prediction of Remission or Activity and Clinical Outcomes in Ulcerative Colitis. Gastroenterology; 2023; 164, pp. 1180-1188.e2. [DOI: https://dx.doi.org/10.1053/j.gastro.2023.02.031]
121. Vande Casteele, N.; Leighton, J.A.; Pasha, S.F.; Cusimano, F.; Mookhoek, A.; Hagen, C.E.; Rosty, C.; Pai, R.K.; Pai, R.K. Utilizing deep learning to analyze whole slide images of colonic biopsies for associations between eosinophil density and clinicopathologic features in active ulcerative colitis. Inflamm. Bowel Dis.; 2022; 28, pp. 539-546. [DOI: https://dx.doi.org/10.1093/ibd/izab122]
122. Ozaki, R.; Kobayashi, T.; Okabayashi, S.; Nakano, M.; Morinaga, S.; Hara, A.; Ohbu, M.; Matsuoka, K.; Toyonaga, T.; Saito, E. et al. Histological risk factors to predict clinical relapse in ulcerative colitis with endoscopically normal mucosa. J. Crohns Colitis; 2018; 12, pp. 1288-1294. [DOI: https://dx.doi.org/10.1093/ecco-jcc/jjy092]
123. Ohara, J.; Nemoto, T.; Maeda, Y.; Ogata, N.; Kudo, S.E.; Yamochi, T. Deep learning-based automated quantification of goblet cell mucus using histological images as a predictor of clinical relapse of ulcerative colitis with endoscopic remission. J. Gastroenterol.; 2022; 57, pp. 962-970. [DOI: https://dx.doi.org/10.1007/s00535-022-01924-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36184701]
124. Iacucci, M.; Jeffery, L.; Acharjee, A.; Grisan, E.; Buda, A.; Nardone, C.M.; Smith, S.C.L.; Labarile, N.; Zardo, D.; Ungar, B. et al. Computer-Aided Imaging Analysis of Probe-Based Confocal Laser Endomicroscopy With Molecular Labeling and Gene Expression Identifies Markers of Response to Biological Therapy in IBD Patients: The Endo-Omics Study. Inflamm. Bowel Dis.; 2023; 29, pp. 1409-1420. [DOI: https://dx.doi.org/10.1093/ibd/izac233]
125. Kellerman, R.; Bleiweiss, A.; Samuel, S.; Margalit-Yehuda, R.; Aflalo, E.; Barzilay, O.; Ben-Horin, S.; Eliakim, R.; Zimlichman, E.; Soffer, S. et al. Spatiotemporal analysis of small bowel capsule endoscopy videos for outcomes prediction in Crohn’s disease. Ther. Adv. Gastroenterol.; 2023; 16, 17562848231172556. [DOI: https://dx.doi.org/10.1177/17562848231172556]
126. Silverman, A.L.; Shung, D.; Stidham, R.E.; Kochhar, G.S.; Iacucci, M. How artificial intelligence will transform clinical care, research, and trials for inflammatory bowel disease. Clin. Gastroenterol. Hepatol.; 2024; 23, pp. 428-439.e4. [DOI: https://dx.doi.org/10.1016/j.cgh.2024.05.048] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38992406]
127. Rimondi, A.; Gottlieb, K.; Despott, E.J.; Iacucci, M.; Murino, A.; Tontini, G.E. Can artificial intelligence replace endoscopists when assessing mucosal healing in ulcerative colitis? A systematic review and diagnostic test accuracy meta-analysis. Dig. Liver Dis.; 2024; 56, pp. 1164-1172. [DOI: https://dx.doi.org/10.1016/j.dld.2023.11.005]
128. Pal, P.; Ramchandani, M.; Patel, R.; Banerjee, R.; Kanaganti, S.; Gupta, R.; Tandan, M.; Reddy, D.N. Role of ultra-high definition endoscopy (endomicroscopy and endocytoscopy) and real-time histologic examination in inflammatory bowel disease: Scoping review. Dig. Endosc.; 2024; 36, pp. 274-289. [DOI: https://dx.doi.org/10.1111/den.14659]
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Inflammatory bowel diseases (IBDs), comprising Crohn’s disease (CD) and ulcerative colitis (UC), are chronic immune-mediated inflammatory diseases of the gastrointestinal (GI) tract with still-elusive etiopathogeneses and an increasing prevalence worldwide. Despite the growing availability of more advanced therapies in the last two decades, there are still a number of unmet needs. For example, the achievement of mucosal healing has been widely demonstrated as a prognostic marker for better outcomes and a reduced risk of dysplasia and cancer; however, the accuracy of endoscopy is crucial for both this aim and the precise and reproducible evaluation of endoscopic activity and the detection of dysplasia. Artificial intelligence (AI) has drastically altered the field of GI studies and is being extensively applied to medical imaging. The utilization of deep learning and pattern recognition can help the operator optimize image classification and lesion segmentation, detect early mucosal abnormalities, and eventually reveal and uncover novel biomarkers with biologic and prognostic value. The role of AI in endoscopy—and potentially also in histology and imaging in the context of IBD—is still at its initial stages but shows promising characteristics that could lead to a better understanding of the complexity and heterogeneity of IBDs, with potential improvements in patient care and outcomes. The initial experience with AI in IBDs has shown its potential value in the differentiation of UC and CD when there is no ileal involvement, reducing the significant amount of time it takes to review videos of capsule endoscopy and improving the inter- and intra-observer variability in endoscopy reports and scoring. In addition, these initial experiences revealed the ability to predict the histologic score index and the presence of dysplasia. Thus, the purpose of this review was to summarize recent advances regarding the application of AI in IBD endoscopy as there is, indeed, increasing evidence suggesting that the integration of AI-based clinical tools will play a crucial role in paving the road to precision medicine in IBDs.
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
Details



1 Unit of Gastroenterology and Digestive Endoscopy, Scientific Institute for Research, Hospitalization and Healthcare Policlinico San Donato, Vita-Salute San Raffaele University, San Donato Milanese, 20097 Milan, Italy; Unit of Gastroenterology and Digestive Endoscopy, Scientific Institute for Research, Hospitalization and Healthcare Policlinico San Donato, San Donato Milanese, 20097 Milan, Italy
2 Unit of Gastroenterology and Digestive Endoscopy, Scientific Institute for Research, Hospitalization and Healthcare Policlinico San Donato, San Donato Milanese, 20097 Milan, Italy
3 School of Specialization in Digestive System Diseases, Faculty of Medicine, University of Pavia, 27100 Pavia, Italy