Digitally Enhanced Plaster of Paris Imaging of Distal Radial Fractures Is Preferred by Orthopaedic Surgeons and Radiologists: A Single Site Survey

Abstract

ABSTRACT

Introduction

X‐rays of bone fractures immobilised with Plaster of Paris (POP) produce images of reduced diagnostic quality due to the increased density and irregular pattern of the POP overlying the anatomy of interest. Post‐processing parameters in digital radiography (DR) can be applied to POP images to increase diagnostic quality without increasing radiation dose. The aim of this study was to evaluate the preferred image quality of POP immobilised distal radius fractures using optimised digital image manipulation algorithms.

Methods

A cross‐sectional, quantitative survey study was conducted between November 2021 and December 2023 at a large metropolitan health network. The manufacturer standard algorithm and three new image post‐processing algorithms were applied to pre‐selected image sets. Orthopaedic surgeons (n = 34) and radiologists (n = 35) were surveyed to rank image quality of 10 random image sets (80 images in total). Data were described and analysed using median rankings, Mann–Whitney U tests and Friedman rank tests with post hoc Wilcoxin rank tests.

Results

A total of 13 orthopaedic surgeons and 14 radiologists participated. A highly enhanced algorithm using contrast boost (Algorithm D) was the most preferred set (n = 18/27) due to better visibility of bony detail and fracture sites. There was no difference in rankings between clinician groups.

Conclusion

In this single site survey, both orthopaedic surgeons and radiologists preferred the highly enhanced post‐processing algorithm (D) indicating that image quality can be improved using optimised digital manipulation. POP post‐processing parameters with contrast boosting could be implemented to potentially increase diagnostic accuracy without increasing radiation dose for x‐ray imaging of the wrist with POP.

Introduction

A target exposure index (EI^T) is an algorithm set by digital radiography (DR) manufacturers to assign an image quality class for x-ray protocols for each body part [1]. The exposure index (EI) value is a measure of the x-ray photons that reach the image receptor and is a marker for exposure quality and deviation from the target EI [2]. The deviation index (DI) value provides feedback to the radiographer of optimal image quality and a corresponding x-ray dose, either above or below the EI^T value [3]. Casting mediums such as Plaster of Paris (POP) and fibreglass casts are a significant cause of variation in DI, which may impair image quality. Casts are used by orthopaedic surgeons to immobilise anatomy in the post fracture setting [4], and due to their increased x-ray absorption, are a significant cause of variation in DI, which may impair image quality [3]. Common practice to compensate for the increased x-ray absorption caused by POP has been to increase the x-ray radiation exposure parameters with subjective variations in x-ray exposure techniques [5, 6].

Plaster casting inherently has an unavoidable artefact projected over the entirety of the area of interest. POP structures may impede the ability to interpret radiographic images due to their similar appearance to bone on x-ray. This is due to general x-ray utilising a stationary flat panel detector and x-ray source to produce two-dimensional images. The absence of volume data means that the overlying layers of objects in the image superimpose one another [1]. Without POP, this is not an issue as with human tissue there is a large enough difference in the linear attenuation of radiation between air, soft tissue and bone that each can be clearly delineated in general x-ray [1]. However, POP and bone are similar enough in density that it becomes difficult to distinguish between the two, which can have the effect of mimicking pathology [7]. Furthermore, there is variability in thickness from one plaster cast to another as well as variations of thickness and density within the same cast [7].

Manufacturers of digital x-ray systems create x-ray imaging protocols with inherent image post-processing algorithms designed to enhance diagnostic quality and interpretation. These algorithms include edge enhancement, noise suppression, improved signal to noise ratio and smoothing [8–10]. It is hypothesised that with development of new optimised post-processing algorithms applied to POP wrist images, improvement of diagnostic image quality may be able to be achieved without the need to increase radiation dose, adhering to the “As low as reasonably achievable” (ALARA) principle [11].

Distal radius fractures are a common pathology in the age group of 50 years and above [12, 13]. The study focuses on this demographic due to the high prevalence of this type of injury. Therefore, the aim of this study was to evaluate the subjective image quality preferences of orthopaedic surgeons and radiologists, in relation to distal radius fractures immobilised with POP using new optimised digital image manipulation algorithms that do not require increasing x-ray exposure.

Methods

Study Design

This cross-sectional, quantitative survey study was conducted in a large metropolitan health network between November 2021 and December 2023. Three new optimised digital image post-processing algorithms were created by the researchers in May 2021 and compared against the standard manufacturer post-processing algorithm. A survey was delivered to two groups: orthopaedic surgeons and radiologists, for their subjective interpretation of image quality. The study is reported here in accordance with the Checklist for Reporting Results of Internet E-Surveys (CHERRIES) [14]. Ethical approval was gained from Eastern Health's Human Research Ethics Committee (reference number LR20-102).

Background Setting and Equipment

The health network in which the study was conducted services a population of approximately 800,000 people and operates three major acute hospital facilities. The health network provides diagnostic imaging and orthopaedic services, employing approximately 34 orthopaedic surgeons and 35 radiologists who provide imaging interpretation services.

General x-ray examinations were performed using Shimadzu RADspeed ceiling mounted x-ray tubes (Shimadzu Corporation, Kyoto, Japan) and generators paired with wireless Canon CXDI flat panel digital image detectors and related CXDI image processing software (Canon, Tokyo, Japan). Data were retrieved from two of the network campuses comprising seven x-ray rooms, each with the same x-ray equipment and detector configuration.

Creation of Optimised Digital Image Algorithms

Post-processing algorithms are created by adjusting variables in Appendix A such as base brightness, base contrast and edge enhancement, etc. For this study, three sets of algorithms were developed for the posterior-to-anterior (PA) and lateral views of the wrist. Algorithm A was pre-determined by the x-ray equipment manufacturer. The post-processing algorithms B, C and D were determined by a senior medical imaging technologist in consultation with the manufacturer application specialist as part of standard quality assurance practice at the hospital (Table 1 and Table 2).

TABLE 1 Post-processing algorithms applied to the PA wrist.

PA wrist	Algorithm A	Algorithm B	Algorithm C	Algorithm D
Manufacturer standard	Intermediate 1	Intermediate 2	Enhanced
Base brightness (Range 1–29)	22	21	19	18
Base contrast (Range 1–29)	15	16	17	18
Edge enhancement (Range 0–20)	10	10	10	10
Edge frequency (Range 1–7)	5	5	5	5
Contrast boost (Range 0–20)	1	4	6	8
Whole region (Range 1–20)	0	5	10	15
Dark region (Range 1–20)	9	10	11	12
Bright region (Range 1–20)	7	7	7	7
Noise reduction (Range 1–10)	5	5	5	5

TABLE 2 Post-processing algorithms applied to the lateral wrist.

Lateral wrist	Algorithm A	Algorithm B	Algorithm C	Algorithm D
Manufacturer standard	Intermediate 1	Intermediate 2	Enhanced
Base brightness (Range 1–29)	22	21	19	18
Base contrast (Range 1–29)	15	15	14	14
Edge enhancement (Range 0–20)	10	11	12	13
Edge frequency (Range 1–7)	5	6	6	7
Contrast boost (Range 0–20)	1	4	8	12
Whole region (Range 1–20)	0	5	10	14
Dark region (Range 1–20)	9	9	10	10
Bright region (Range 1–20)	7	9	11	12
Noise reduction (Range 1–10)	5	5	5	5

The new optimised post-processing algorithms applied to POP images aimed to: increase the detectability of objects; increase differentiation of greyscales between objects; improve the contrast of the POP/bone region; and reduce darkness to balance the whole image quality (Appendix A).

Selection of Sample Wrist Images

A retrospective audit was performed of 50 randomly selected full-cast POP radius x-ray studies referred to the medical imaging department by emergency and orthopaedics. The 50 randomised studies were checked to ensure that the manufacturer standard wrist exposure protocol had been applied (Table 3). Images for inclusion were of people aged over 50 years who had PA and lateral wrist x-rays with POP located in the Picture Archiving and Communication System (PACS) between 1st April 2021 and 25th April 2021. This demographic group was chosen as distal radius fractures are a common injury and possess a lower bone density than a younger demographic [12, 13]. This also allowed for a greater sample size. Each study had 1 PA and 1 lateral image of the wrist. The x-ray studies were entered into an Excel form with identification numbers manually assigned to each study, and an Excel randomisation formula was applied. A total of 10 image sets were randomly selected from the original 50 full-cast POP distal radius x-ray studies. Four post-processing algorithms were applied to each of the PA and lateral views (Figure 1) from the ten selected studies, creating eighty images for the survey. A research accession number was created and the images sent to PACS: Algorithm A (standard manufacturer), B (intermediate 1), C (intermediate 2) and D (enhanced) (Tables 1 and 2).

TABLE 3 Manufacturer standard non-Plaster of Paris (POP) wrist x-ray exposure factors as per vendor recommendations.

Protocol	kVp	mA	Exposure time (ms)	mAs
PA wrist	52	100	16	1.6
Lateral wrist	55	100	20	2.0

[IMAGE OMITTED. SEE PDF]

Survey Participants

A convenience sample of orthopaedic surgeons (n = 34) and radiologists (n = 35) employed by the health service was invited to participate. Information about the study was distributed via the directors of orthopaedics and radiology throughout the health network. Consent was implied by the return of the completed survey.

Survey

A survey (Appendix B) was created by medical imaging interns with consultation from tutor radiographers and researchers (PK, HS). Four radiographers trialled the survey, resulting in minor wording changes to improve clarity. The survey asked orthopaedic surgeons and radiologists to rate their preference for the post-processing algorithm in terms of image quality and, therefore, diagnostic value. A final open-ended question allowed for optional comments for participants to clarify their answers. The survey was created in both an online (Qualtrics, Provo, UT) and paper-based format. The 80 images were retrieved from the radiology PACS system by their assigned accession numbers and displayed on image viewing platforms routinely used by the participants.

The 10 image sets (lateral and PA wrists) were each presented with the application of the 4 post-processing algorithms (A, B, C, D) in a face-to-face meeting with the radiologists and, due to COVID-19 restrictions, in an online format to the orthopaedic surgeons. The survey participants were instructed not to alter the contrast and brightness of the images and to provide a diagnostic opinion of the images as presented. Radiologists viewed the images on high-resolution diagnostic monitors in a small, darkened office, typical for their usual practice. The online image sets were reviewed by orthopaedic surgeons in the typical viewing conditions and environment: a standard rear-projector screen in a spacious, semi-dimmed conference room. Apart from the location and viewing platform, other aspects of the survey procedure were the same for both groups of participants.

The images were displayed to the survey participants on PACS in a 2 × 2 format. This display format allowed the reviewer to view the same PA wrist image with all four post-processing algorithms applied. This was repeated for the lateral wrist views. The order of the images from each image set was manually randomised to avoid the display of an applied algorithm occurring in the same order as the previous image set and to reduce the effect of pattern recognition. Participants were asked to rank each image set from most (rank = 1) to least preferred (rank = 4) with the option to provide a reason for their preference (Appendix B). All survey results were entered into Qualtrics for data analysis.

Analysis

Analysis was completed in two phases using SPSS v 29.0. As data were expected to be non-parametric, median rankings were obtained for the two groups (radiologists and orthopaedic surgeons) for each of the 10 wrist studies. Mann–Whitney U tests were conducted to determine if there was a difference in median rankings between the two groups for both the PA and lateral views. The ranking of images for each clinical group was described using frequencies and proportions, and data were also visually inspected using bar charts to determine the most preferred algorithm.

Data were combined, and Friedman's rank tests with post hoc Wilcoxin rank tests with Bonferroni adjustment for multiple comparisons were applied when no differences in algorithm preference were observed between clinical groups. This was completed to determine the order of ranking of algorithm preference overall.

Open ended responses were analysed using summative content analysis to identify patterns in the data.

Results

Thirteen orthopaedic surgeons completed the survey, representing approximately 38% of the total number of orthopaedic surgeons from the health network. Fourteen (34%) radiologists completed the survey. Most clinicians were male (n = 9, 64%) and experienced consultants with an average of 13 years (range 1–40) experience.

Image Algorithm Preferences

There was no difference in ratings between the radiologists and the orthopaedic surgeons (Table 4, Appendix C). Overall, 18 participants selected the enhanced algorithm D as the most preferred digital image filter combination (n = 18/27) (Figure 2). Appendix D provides ranking of algorithm preference using the Friedman rank test.

TABLE 4 Median difference in preferences between each profession for each algorithm applied to Plaster of Paris (POP) images.

	Radiologists	Orthopaedic surgeons	Mann Whitney U, p value
Median (IQR)	Number ranked as 1 (%)	Median (IQR)	Number ranked as 1 (%)
Algorithm A PA view	4 (4–4)	6 (2)	4 (4–4)	5 (2)	50, p = 1.000
Algorithm A lateral view	4 (4–4)	4 (1)	4 (4–4)	3 (1)	50, p = 1.000
Algorithm B PA view	3 (3–3)	10 (3)	3 (2.75–3)	18 (7)	40, p = 0.481
Algorithm B lateral view	3 (3–3)	11 (4)	3 (2.75–3)	20 (8)	40, p = 0.481
Algorithm C PA view	2 (2–2)	15 (5)	2 (2–2)	34 (13)	50, p = 1.000
Algorithm C lateral view	2 (2–2)	36 (13)	2 (2–2)	33 (12)	50, p = 1.000
Algorithm D PA view	1 (1–1)	109 (40)	1 (1–1)	73 (27)	55, p = 0.739
Algorithm D lateral view	1 (1–1)	89 (32)	1 (1–1.125)	74 (28)	51, p = 1.000

[IMAGE OMITTED. SEE PDF]

There was a statistically significant difference in the image set preferences of clinicians for both PA (χ² (3) = 30, p < 0.001) and lateral views (χ² (3) = 30, p < 0.001). Algorithm D (highest contrast filter) was preferred for both PA and lateral wrist views (Z = −3.162, p = 0.002) (Figure 2). The manufacturer standard algorithm A was least preferred for both PA and lateral views.

Reasons for Image Preference

Open-ended survey responses (Appendix E) demonstrated that both orthopaedic surgeons and radiologists were primarily interested in the alignment of anatomy when choosing their preferred images. The main reason for the selection of Algorithm D for both participant groups was due to greater image contrast providing better visualisation of bony detail, fracture characterisation and bony alignment. Radiologists and orthopaedic surgeons also reported looking for evidence of fracture lines as a common reason for their preference. Orthopaedic surgeons commented on the images for the evaluation of the POP moulding. Survey participants recognised that digital enhancement of POP could be applied to other anatomical regions such as elbow, knee and ankle.

Discussion

This study evaluated the quality of images manipulated by digital post-processing algorithms for POP wrist x-ray images. Balancing diagnostic accuracy and radiation dose is crucial to minimise radiation exposure while maintaining image quality. It is important to consider the value of customising algorithms to meet the needs and preferences of end users such as orthopaedic surgeons and radiologists when developing and implementing new POP imaging protocols for wrist fractures and other anatomical areas. Orthopaedic surgeons and radiologists preferred a new optimised, enhanced algorithm (D) for both PA and lateral views. The combination of increasing the detectability of objects, increasing differentiation of greyscales between objects, improving the contrast of the POP/bone region, and reducing darkness to balance the whole image quality were used to enhance image quality. The net effect resulted in higher subjective image quality. Subjective image quality increased from algorithm A to D as the intent of the algorithm was to increase detectability of objects, differentiation of greyscales between objects, and the contrast of the POP/bone regions with each algorithm change. No major differences in image preference were observed between the two clinician groups.

This study provides evidence that digital image processing can improve subjective image quality while maintaining the same exposure factors. A potential positive consequence of this is that the patient with a POP is not exposed to additional radiation as may occur with the common practice of enhancing imaging quality by increasing x-ray exposure parameters. Adjusting post-processing algorithms is an example of how radiographers can hypothetically manipulate images and maintain the ALARA principle to ensure optimised patient radiation dose [11]. To our knowledge, no studies have formally evaluated the use of digital post-processing to enhance x-ray imaging of fractures with the application of POP; therefore, these findings offer a novel area for future research.

Overall, image preferences were similar between the clinical groups. This is noteworthy given the important role both radiologists and orthopaedic surgeons play in the management of distal radius fractures. Reasons for their image preferences were also broadly similar. However, orthopaedic surgeons and radiologists differed in what features of the images were considered important in their interpretation. Both groups noted the importance of correct radiographic positioning of the anatomy, allowing for better evaluation of the anatomical alignment. Radiologists and orthopaedic surgeons were commonly concerned with visualisation of fracture lines. Orthopaedic surgeons were also interested in the application of POP moulding for evaluation of correct immobilisation of the anatomy. Consensus among clinician groups reinforces the possibility of applying digitally enhanced image filters to other anatomical regions such as ankles, knees or elbows to aid diagnostic interpretation.

Strengths

This study is the first to survey orthopaedic surgeons and radiologists in their preferences of post processing algorithms applied to x-ray images of the wrist in POP. The study was also conducted in a large public hospital setting to aid generalisability. Distal radius fractures are common in the age group of 50 years and above, and there is potential for a reduction of the x-ray radiation burden of this demographic. This can also be applied to any demographic requiring x-ray examination with POP immobilisation.

Limitations

Variations in the survey format due to COVID-19, the image viewing platform and environment between the two participant groups may have resulted in additional variation to the images viewed by clinicians. However, in this pragmatic study, these variations reflect real-life clinical practice and did not result in significant differences in viewing between the two clinical groups. Orthopaedic surgeons and radiologists may also be influenced by their experience and learned pattern recognition in the assessment of image sets as only four different image algorithms were produced. However, randomisation of images was performed to minimise this issue. Presenting data to a panel of expert radiographers may have also been considered. We made an assumption that the presented images were clinically acceptable, but this was not tested. Sample size was insufficient to perform further analysis on the impact of experience; however, most clinicians had similar levels of experience. Variation in fracture patterns and whether the plaster was wet or dry may also have influenced results, but it was beyond the scope of this study to evaluate these variables. However, open-ended responses relating to the reasons for participant preference indicated contrast was the main reason for image preference to better visualise bony detail, alignment and fracture characterisation. It should also be acknowledged that POP is not the only casting medium used in clinical practice, which may have different implications on imaging parameters. This is a possible area for future research, as is the evaluation of plaster moulding.

Conclusion

This study demonstrated that digital image post-processing enhancement algorithms can improve diagnostic quality and interpretation of POP wrist x-ray images. Orthopaedic surgeons and radiologists have similar image preferences in relation to digitally processed images. The digital image post-processing technology has the potential to be applied to other anatomical regions with the application of POP. The study may influence and change the common practice of increasing x-ray radiation exposure for POP imaging and reduce the community x-ray radiation burden in accordance with the overriding principle of ALARA.

Acknowledgements

We would like to thank Carina Mojet, Barbara Nourish, Helen Johnston, Edward Crawford, Simon Mai, Chris Choi and Cheryl Li for their assistance in data collection and analysis. We acknowledge Shimadzu Medical Systems Australia for assisting in the creation of the new post-processing algorithms and the Radiologist provider, Imaging Associates Radiology. Open access publishing facilitated by La Trobe University, as part of the Wiley - La Trobe University agreement via the Council of Australian University Librarians.

Ethics Statement

Ethical approval was gained from Eastern Health's Human Research Ethics Committee (reference number LR 20–102).

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

J. A. Seibert, “Projection X‐Ray Imaging: Radiography, Mammography, Fluoroscopy,” Health Physics 116, no. 2 (2019): 148–156.

H. Park, Y. Yoon, J. Kim, et al., “Use of Clinical Exposure Index and Deviation Index Based on National Diagnostic Reference Level as Dose‐Optimization Tools for General Radiography in Korea,” Radiation Protection Dosimetry 191, no. 4 (2020): 439–451.

J. Guðjónsdóttir, K. Paalsson, and G. Sveinsdóttir, “Are the Target Exposure Index and Deviation Index Used Efficiently?,” Radiography 27, no. 3 (2021): 903–907.

H. Sharma and D. Prabu, “Plaster of Paris: Past, Present and Future,” Journal of Clinical Orthopaedics and Trauma 4, no. 3 (2013): 107.

J. Lampignano and L. E. Kendrick, Bontrager's Textbook of Radiographic Positioning and Related Anatomy‐E‐Book (Elsevier Health Sciences, 2017).

S. P. Knight, “A Paediatric X‐Ray Exposure Chart,” Journal of Medical Radiation Sciences 61, no. 3 (2014): 191–201.

D. E. Whittier, S. L. Manske, S. K. Boyd, and P. S. Schneider, “The Correction of Systematic Error due to Plaster and Fiberglass Casts on HR‐pQCT Bone Parameters Measured In Vivo at the Distal Radius,” Journal of Clinical Densitometry 22, no. 3 (2019): 401–408.

Y. Abdallah and R. Yousef, “Augmentation of X‐Rays Images Using Pixel Intensity Values Adjustments,” International Journal of Science and Research (IJSR) 4, no. 2 (2015): 2425–2430.

A. K. Bharodiya and A. M. Gonsai, “An Improved Edge Detection Algorithm for X‐Ray Images Based on the Statistical Range,” Heliyon 5, no. 10 (2019): e02743.

M. S. Lee, C. H. Park, and M. G. Kang, “Edge Enhancement Algorithm for Low‐Dose X‐Ray Fluoroscopic Imaging,” Computer Methods and Programs in Biomedicine 152 (2017): 45–52.

D. J. Kaplan, J. N. Patel, F. A. Liporace, and R. S. Yoon, “Intraoperative Radiation Safety in Orthopaedics: A Review of the ALARA (As Low as Reasonably Achievable) Principle,” Patient Safety in Surgery 10 (2016): 1–7.

N. J. MacIntyre and N. Dewan, “Epidemiology of Distal Radius Fractures and Factors Predicting Risk and Prognosis,” Journal of Hand Therapy 29, no. 2 (2016): 136–145, https://doi.org/10.1016/j.jht.2016.03.003.

A. Azad, H. P. Kang, R. K. Alluri, V. Vakhshori, H. F. Kay, and A. Ghiassi, “Epidemiological and Treatment Trends of Distal Radius Fractures Across Multiple Age Groups,” Journal of Wrist Surgery 8, no. 4 (2019): 305–311.

G. Eysenbach, “Improving the Quality of Web Surveys: the Checklist for Reporting Results of Internet E‐Surveys (CHERRIES),” Journal of Medical Internet Research 6 (2004): e34.