Quantitative MRI relaxometry in brain tumor needle biopsies: Multimodal comparison with tissue fluorescence, radiology, and neuropathology

Abstract

Background

Quantitative MRI (qMRI) relaxometry holds potential for brain tumor identification beyond contrast enhancement on conventional images. However, clinical implementation is limited by long acquisition times, changing conditions between imaging and surgery, and lack of correlation with standard techniques.

Purpose

To extend a methodology for multimodal data analysis to relaxometry data. To integrate relaxometry into the burr hole needle biopsy procedure with optical guidance, setup a workflow for multimodal data processing and analysis, and apply the methodology in a clinical setting.

Methods

Multi-dimensional multi-echo relaxometry data (2x6 min) was acquired in addition to the clinical imaging protocol. Relaxation rate and proton density maps, as well as their differences were calculated before (R₁, R₂) and after gadolinium contrast-agent administration (R₁Gd, R₂Gd). Radiological volumes of interest (VOIs: tumor, edema, white matter, and biopsy) were defined on clinical images. Rate distribution changes were analyzed on three levels: the biopsied volume, along the needle trajectory (4x4x4 mm³ volumes), and VOIs. Increased R₁Gd and R₂Gd were compared to indications from 5-aminolevulinic acid-induced fluorescence and detailed neuropathological evaluation.

Results

Neuropathological analysis confirmed seven glioblastoma, one lymphoma, and one non-tumorous diagnosis. Increased R₁Gd was found in all biopsied volumes, although tumorous volumes presented larger R₁Gd increase (3–9 times) compared to volumes dominated by necrotic or non-tumorous tissue. Along the trajectory, increased R₁Gd and R₂Gd were not tumor-specific, however, the greatest R₁Gd shifts were found in or adjacent to radiologically defined tumorous tissue. Increased relaxation rates corresponded to 82% and 45% (R₁Gd: φ = 0.35, R₂Gd: φ = 0.27) of fluorescence peaks. In the radiological VOIs, increased R₁Gd and R₂Gd were found in tumorous tissue, a slight right shift in edematous tissue, and negligible changes in white matter.

Conclusion

Combined analysis suggests increased R₁Gd together with fluorescence peaks as a marker for tumor tissue. The presented multimodal approach provides a workflow toward clinical translation of relaxometry.

Introduction

Brain tumor diagnoses rely on representative tissue samples taken during, e.g., a needle biopsy procedure. Due to the heterogeneous nature of high-grade brain tumors, finding representative samples is a challenging task [1,2].

Imaging techniques such as quantitative MRI (qMRI) hold potential for non-invasive tissue characterization by relating the measured signal to underlying physical properties [3,4] represented by quantitative maps. Relaxation rate maps are derived from a signal fit over multiple echo and delay times [5], and depict e.g., relaxation time (T) or rate (R = 1/T). Even though relaxometry was studied already in the 1970s, advancements in hardware and computing power have renewed its potential, indicating pathology in multiple sclerosis, epilepsy [6], and brain tumors. In two recent papers, a difference in T or R relative to administration of gadolinium-based (Gd) contrast media has been suggested as support for brain tumor tissue identification and infiltration beyond contrast enhancement on conventional MRI [7,8]. Blystad and colleagues assessed longitudinal relaxation rate (R₁) difference before and after Gd contrast administration and demonstrated enhancement in the peritumoral zone in glioma patients [7]. In addition, Maurer et al. found a shortened longitudinal relaxation time (T₁) after contrast in tumorous tissue for patients with isocitrate dehydrogenase (IDH) wildtype gliomas [8]. Although promising, extensive acquisition times have hampered the clinical pertinence of relaxometry. Moreover, changes between image acquisition and surgery limit the reliability of correlation studies between relaxometry data and standard techniques. These changes include patient positioning, registration errors during neuronavigation, movement as equipment is fastened, and altered intracranial conditions as the skull and dura are opened.

To overcome the limitations of preoperative imaging and identify tumorous tissue during surgery, intraoperative techniques have been introduced [9,10]. Previously, a system for optical intraoperative feedback of 5-ALA-induced tissue fluorescence through protoporphyrin IX (PpIX) accumulation associated with high-grade glioma and lymphoma [11,12] has been developed [13]. The system offers real-time measurements before tissue sampling through a probe-based system. Recently, the system was integrated into the surgical procedure [14] and pre-, intra-, and postoperative data combined for multimodal analysis. While intraoperative techniques allow feedback in situ, the possible adjustments are confined to the planned trajectory. Hence, there is still a need to improve representative tissue identification in target planning.

Therefore, the overall aim of this study was to extend the methodology for multimodal data analysis of pre-, intra-, and post-operative data to relaxometry on the millimeter scale. The specific aims were to integrate a relaxometry sequence suitable for clinical practice in the image acquisition protocol for brain tumor patients and set up a pipeline for registration to multimodal data, allowing comparison of relaxometry findings to fluorescence along the trajectory and neuropathological parameters in the biopsied volume. The methodology was applied in patients undergoing brain tumor needle biopsy surgery.

Materials and methods

Patients

Ten patients referred for brain tumor needle biopsy were included in the study (mean age 64 years, range 18–79 years, two women). The Ethical Review Authority granted study approval (2020–01404), and all patients gave written informed consent before inclusion (inclusion period: Jan 1^st, 2022, to Aug 31^st, 2023). In line with the clinical protocol for fluorescence-guided resection [12], a standard oral dose of Gliolan^® (20 mg/kg, Medac GmbH, Wedel, Germany) was given to each patient 2–3 hours before general anesthesia. One patient was excluded from analysis since the needle trajectory crossed the tentorium cerebelli, thus probing tissue outside the scope of relaxometry analysis. The optical data have previously been reported in a methodology study on the workflow for optical integration [14] and a 20-patient study of optical trends [15].

Data acquisition

The multimodal data acquisition, including preoperative conventional imaging, relaxation rate map derivation, intraoperative optical measurements, and neuropathological diagnosis, is depicted in Fig 1.

[Figure omitted. See PDF.]

(A) Radiology: clinical image acquisition (T₁w gadolinium (Gd) contrast-enhanced and T₂wFLAIR), calculated quantitative maps (R₁, R₂, PD before and after Gd contrast-enhancement), and example segmentation of radiological volumes of interest (VOI). (B) Surgical set up with neuronavigation and optical guidance, followed by tissue sampling and neuropathological evaluation of a Hematoxylin and Eosin-stained slide.

Radiology

MRI was acquired on a 3T scanner (Skyra or Prisma, depending on availability, Siemens Healthineers, Erlangen, Germany) within four days before surgery using a 20-channel head coil. The clinical brain tumor imaging protocol included conventional T₁-weighted (w), T₂w, T₂w-FLAIR, and T₁wGd contrast-enhanced images (Gd dose: 16–17 mL, Dotarem, 279 mg/mL, Gothia Medical, Sweden). A qMRI multi-dynamic multi-echo (MDME) sequence [5] was added before and after Gd administration, allowing calculation of several image contrasts from a single scan, see acquisition details in Table 1. Within 24 hours of surgery, a postoperative computed tomography (CT, SOMATOM Definition Edge, Siemens) or T₁w MRI (3T Skyra, Siemens; 3T Ingenia or 1.5T Achieva, Philips, Best, The Netherlands; 1.5T Optima, GE Healthcare, Chicago, USA) was acquired.

[Figure omitted. See PDF.]

Synthesized conventional images (synT₁w, synT₂w, synT₁wGd, synT₂wGd), relaxation time, and proton density maps (T₁, T₁Gd, T₂, T₂Gd, PD) were derived from the qMRI MDME sequence through the SyMRI software (v0.45.38, SyntheticMR AB, Linköping, Sweden) and relaxation rate maps (R₁, R₁Gd, R₂, R₂Gd) were calculated.

Radiological volumes of interest (VOIs) were defined by a senior neuroradiologist (I.B.) on preoperative conventional MR images in 3D Slicer [16]. Identified volumes were gross tumor volume (GTV) and peritumoral edema (PTE) in relation to the biopsy needle trajectory. The GTV was specified as the contrast-enhancing tumor on T₁wGd, including the necrotic core and tumorous tissue in the T₂w volume. PTE was identified as the hyperintense region on T₂wFLAIR surrounding the GTV. Reference volumes of contralateral normal-appearing white matter (NAWM) VOIs (∅ 1 cm) were defined in axial slices on par with the lateral ventricle in the frontal and parietal lobes. In two patients, bilateral ischemic changes were present, thus, the NAWM VOI diameters were reduced (∅ 0.5 cm).

Surgery

An in-house developed probe-based spectroscopy system [14,17] was used for intraoperative measurements of tissue fluorescence. In brief, tissue is excited by a near-UV laser, and three consecutive spectra within the wavelength (λ) interval 450–750 nm are captured and presented in real-time. The spectrum contains autofluorescence and, in the case of protoporphyrin IX (PpIX) accumulation, a peak at λ = 635 nm. Measurements were integrated into the surgical procedure by fitting the optical probe into the outer cannula of a biopsy needle (sampling window: 8 mm, Passive Biopsy Needle Kit, Medtronic Inc., MN, USA). In turn, the outer cannula was modified with an aperture at the tip to allow forward-looking measurements.

The needle biopsy procedure with optical guidance is described in detail in [14]. Briefly, a trajectory and target (defined as 0 mm) were planned on preoperative T₁wGd or T₂wFLAIR images. Images were registered to the patient’s anatomy with a neuronavigation system (StealthStation S8, Medtronic Inc.), and errors (registration and target) were noted. The skull and dura were opened (3.2–10 mm), and the AutoGuide^® (Medtronic Inc.) held the probe-cannula kit in place during manual insertion toward the target (step size: 1–13 millimeters). MRI coordinates and tissue fluorescence spectra were acquired in each position along the trajectory. The highest PpIX peak near the planned target was identified, and the biopsy cannula window was adjusted to overlap the specified area. Keeping the outer cannula in place, the probe was retracted, the inner cannula inserted, and tissue samples were taken in four directions, separated by a 90-degree needle rotation in the plane orthogonal to the trajectory. If the surgeon deemed the tissue material too sparse, another set of rotations or a second position was sampled. Tissue samples were sent for intraoperative smear analysis. After receiving a preliminary diagnosis from the neuropathologist, surgery was closed. Final diagnoses were based on the 2021 WHO classification of CNS tumors [18].

The formalin-fixed material was divided into 3 μm sections and stained with Hematoxylin and Eosin (H&E) and Ki67. Molecular analyses included O⁶-methylguanine-DNA-methyltransferase (MGMT) methylation (Therascreen MGMT Pyro kit, QIAGEN, Germany), IDH mutation (Therascreen IDH1/2 RGQ PCR kit, QIAGEN), and 1p19q codeletion. In addition to the clinical diagnosis, a senior neuropathologist (M.H.) re-evaluated the digitized stained slides. The re-evaluation assessed the morphological heterogeneity of the samples; presence and percentage of high- and low-grade tumor cells, non-tumor tissue, and necrosis.

Data processing

The processing pipeline was built in Python (v.3.11) utilizing the Nipype framework [19] including image corrections with FMRIB’s Software Library [20] and co-registration with Advanced Normalization Tools (ANTs [21]). Detailed steps of the processing pipeline are illustrated in S1 Fig in the Supplemental Material.

Following bias field correction and skull-stripping, the conventional images were co-registered to T₁wGd space using a rigid (preoperative) or affine (postoperative) transform. Additionally, the transform between the navigation system and native T₁wGd space was calculated. A rigid transform was calculated between synT₂w and synT₂wGd space, followed by transform calculation between synT₁wGd and T₁wGd space (ANTs). GTV, PTE, and NAWM VOIs were transformed into native qMRI spaces for analysis.

The final entry and biopsy positions were identified on postoperative imaging. The entry point was defined by the trephination made in the skull (on CT) or opening in the dura (on MRI). The hypointense visible impact from trajectory and tissue sampling in the postoperative image was used to define the final biopsy position and volume. Measurement coordinates from the navigation system were aligned to the final trajectory by Euclidean distance. A volume (V_Traj, 4x4x4 mm³; 25–37 voxels) was defined in each measurement position along the final trajectory. V_Traj and biopsy volumes were confirmed and classified according to tissue type (GTV, PTE, gray matter (GM), or WM) on conventional images by a senior neuroradiologist (I.B.). Volumes containing several tissue types, e.g., PTE and GTV, or GTV and WM, were identified.

Fluorescence signals were represented as a ratio between the maximum local intensity in the 633–637 nm range (corrected for autofluorescence contribution) and the maximum autofluorescence intensity around 500 nm. After disregarding noise-dominated spectra (signal-to-noise ratio 5:1), a PpIX peak was defined as a ratio >0.1.

Data analysis and statistics

The translational displacement between qMRI acquisitions before and after Gd administration was quantified through the L2 norm. The Euclidean distance (median, interquartile range (IQR)) between planned and final trajectory positions was compared to the registration and target errors reported in the navigation system. Subsequent multimodal analysis was divided into the biopsied tissue volume, trajectory, and radiological VOI findings.

In the final biopsy volume, R₁, R₂, R₁Gd, and R₂Gd distribution changes were assessed through a two-sided Mann-Whitney-U-test corrected for Type 1 error of multiple comparisons using False discovery rate (FDR) with the Benjamini-Hochberg method [22] (p-values <0.05 were considered significant). Rate increases were investigated in relation to PpIX peak presence, tumor percentage, and Ki67 index.

Along the trajectory, R₁, R₂, R₁Gd, and R₂Gd in each V_Traj were compared with a two-sided Mann-Whitney-U-test using FDR or multiple comparison correction (corrected p-value <0.05). Inter-modality correlations between R₁, R₂, R₁Gd, and R₂Gd V_Traj distributions, radiological classification on conventional imaging, and fluorescence were assessed via the Phi coefficient (φ). Per tissue type density heatmaps of R₁, R₂, R₁Gd, and R₂Gd were investigated and compared to the corresponding VOI values.

Finally, the median, IQR, unbiased skew, and voxel count for R₁, R₂, R₁Gd, and R₂Gd, and PD distributions in the radiological VOIs (GTV, PTE, NAWM) were reported.

Results

A qMRI relaxometry protocol was integrated into clinical practice, adding 2x6 min to the imaging acquisition time. The translational displacement between pre- and postoperative imaging was 0.42–6.3 mm. The median Euclidean distance between the planned and final trajectory was 2.8 mm (1.6–7.0 mm) while the intraoperative registration and target errors in the navigation system were 1.5  mm (1.5-1.6 mm) and 0.4  mm (0.2-0.5 mm). Neuropathological diagnoses were Glioblastoma IDH-wildtype, CNS WHO grade 4 (Patients 1–7), Primary diffuse large B-cell Lymphoma (Patient 8), and non-tumor (Patient 9). A noticeable increase in astrocytes was found for the non-tumor sample, but no indication of malignancy.

Biopsied tissue volume

Neuropathological data, relaxation rates, and PpIX peak occurrence are presented in Table 2. Overall, increased R₁Gd was observed in seven tumorous and non-tumorous biopsy volumes, and R₂Gd was increased in four tumorous samples. Relaxation rates showed a large variability within and between patients. PpIX peaks were found in tumorous biopsy volumes but not in the non-tumorous samples. The neuropathological re-evaluation revealed that two of the tumorous samples contained a majority of non-tumor tissue (Patient 1) or necrosis (Patient 7).

[Figure omitted. See PDF.]

Joint assessment showed agreement between increased R₁Gd, PpIX peaks, and tumor diagnoses in seven patients, while Patient 7 presented decreased R₁Gd and PpIX peaks, see Fig 2. Tumorous biopsy volumes had a larger increase in R₁Gd than necrotic or non-tumorous volumes (3–9 times). Increased R₂Gd and PpIX peaks were found in the biopsy volume of four patients, and decreased R₂Gd in one patient. Interestingly, PD percentages included values >100% and showed variable trends after Gd administration compared to before (Table 2: five increased, two unchanged, two decreased).

[Figure omitted. See PDF.]

Patients 1-7: glioblastoma (blue), Patient 8: lymphoma (turquoise), Patient 9: non-tumor (black). The dashed lines denote the population median, upper, and lower quartiles. (B) Fluorescence response with or without PpIX peak. (C) H&E-stained frozen section from digital pathology PACS. The black horizontal lines indicate 100 μm.

Along the trajectory

The multimodal needle trajectory data per patient are presented in Fig 3. Along the needle trajectory (V_Traj = 75), radiological definition of tumor (GTV) spanned −18 mm to +10 mm of the planned target (Fig 3A). Increased R₁Gd was observed in all tissue types, along the whole trajectory (φ: 0.39, Fig 3B). R₂Gd increases were found in all but PTE tissues, also covering the whole trajectory (φ: 0.31, Fig 3B). In agreement with radiological definition of GTV, PpIX fluorescence peaks ranged from −18 mm to +10 mm (φ: 0.61, Fig 3C).

[Figure omitted. See PDF.]

Left column: Radiological classification, R₁ relaxation rate difference, and fluorescence peaks in relation to distance from planned target position (0 mm). Right column: Data examples. A) Radiological classification of measurement volume into GM, GTV, PTE, or WM on clinical imaging. (B) Longitudinal (R₁, black) and transverse (R₂, gray) relaxation rates after Gd compared to before Gd contrast-agent administration (corrected p-value < 0.001). Filled lines indicate an increase, and hollowed-out lines indicate a decrease. C) PpIX-fluorescence peak positions, i.e., where fluorescence ratio >0.1. Gray boxes correspond to the needle biopsy volumes (8 mm) defined by the final biopsy position on postoperative imaging. The larger volumes for patients 3, 8, and 9 are due to the surgeon deeming the material at the first biopsy position too sparse. Values in parenthesis specifies the total number of volumes for that tissue and data type. Gd: gadolinium, GM: gray matter, GTV: gross tumor volume, PpIX: protoporphyrin IX, PTE: peritumoral edema, WM: white matter.

Combined analysis suggested an overlap between PpIX peaks and increased R₁Gd (φ: 0.35, 18/22 V_Traj). These V_Traj corresponded to the radiological definition of GTV but also included PTE (Patient 1). The four V_Traj where PpIX fluorescence but no increased R₁Gd was found were classified as GTV and WM on conventional imaging (Patients 2, 7 and 8). Four GTV V_Traj did not exhibit any R₁Gd increase or PpIX fluorescence (Patients 1, 7 and 8). The overlap between PpIX peaks and increased R₂Gd was lower than that of R₁Gd and confined to V_Traj defined as GTV (φ: 0.27, 10/22 V_Traj). Per patient, per V_Traj rate distributions together with the fluorescence ratio are presented in S2 and S3 Figs in the Supplemental Material.

GTV V_Traj showed greater R₁ and R₂ value dispersion than the other tissue types (Fig 4A-D), ranging from the cerebrospinal fluid-to-GM axis (values similar to PTE) to values greater than that of WM. Additionally, a trend of greater R₁Gd shifts in GTV V_Traj than GM, PTE, and WM was observed (Fig 4E-H, Table 3).

[Figure omitted. See PDF.]

[Figure omitted. See PDF.]

(A-D) Per tissue distribution before and (E-H) after gadolinium (Gd) contrast enhancement. (A, E) Gray matter (GM). (B, F) Gross tumor volume (GTV). (C, G) Peritumoral edema (PTE), and (D, H) White matter (WM). Patient 9 had no radiological definition of GTV or PTE. Circles depict median and IQR before (dotted line) and after administration of Gd (solid line).

Radiological volumes of interest

General trends of lower R₁ and R₂ in GTV compared to NAWM were indicated in the density heatmaps (Fig 5, Table 4). Additionally, increased R₁Gd was observed in the GTV, a slight right shift in the PTE VOI, and constant values in the NAWM VOI (Table 4).

[Figure omitted. See PDF.]

[Figure omitted. See PDF.]

(A-C) Per tissue distribution before and (D-F) after gadolinium (Gd) contrast enhancement. (A, D) Gross tumor volume (GTV). (B, F) Peritumoral edema (PTE), and (C, F) Normal appearing white matter (NAWM). Patient 9 had no radiological definition of GTV or PTE. Circles depict median and IQR before (dotted line) and after administration of Gd (solid line).

Compared to the relaxation rates in the VOIs, median R₁ along the trajectory was higher for GTV and lower for PTE and WM. Median R₂ was skewed toward higher values for GTV and lower for PTE. For WM, the values were comparable. An increase in PDGd percentage was found for GTV. Median R₁ rates in the biopsied volumes were lower than those of NAWM for eight patients and similar to NAWM for one patient (Patient 7).

Discussion

Reliable identification of representative tissue for tumor diagnosis would be exceptionally valuable for the clinical treatment of brain tumors. The present study added qMRI relaxometry with clinically acceptable acquisition times to the routine imaging protocol and a processing pipeline for multimodal brain tumor biopsy data analysis. Multimodal analysis indicated that increased R₁Gd and R₂Gd values were not tumor-specific, but R₁Gd increase was greater in tumorous tissue compared to non-tumorous or necrotic tissue, as confirmed by neuropathology. Increased R₁Gd was also found in 82% of PpIX peak positions. To our knowledge, this is the first study to compare relaxometry to multimodal pre-, intra-, and post-operative data on the millimeter scale.

Relaxation rates after gadolinium contrast enhancement

In the gross tumor volume.

A general trend of increased R₁Gd (T₁ shortening) was found in tissue radiologically defined as GTV, as corroborated by other studies [7,8,23,24]. The increase indicates interaction of Gd with tissue outside vessels, signifying a leaky blood-brain barrier [8,25]. The relaxation rate increase is proportional to tissue Gd concentration [7,26]. Although longitudinal relaxation effects (R₁) of Gd tend to dominate [27], a slight right shift in R₂Gd was seen in the radiological GTV VOI, along the trajectory in V_Traj classified as GTV and WM, and in five biopsy volumes.

Beyond the gross tumor volume.

Outside the radiologically defined tumor volume, a slight right shift in R₁Gd was found for the PTE VOI. Blystad and colleagues presented increased R₁Gd in PTE close to GTV, indicating enhancement beyond what is seen on conventional imaging [7]. In the present study, four WM and PTE dominant V_Traj with increased R₁Gd were found close to the radiological definition of GTV. Interestingly, four V_Traj close to GTV showed decreased R₁Gd (GTV, PTE, and GM). The altered rates could be due to edema, noise, or potentially, partial volume effects arising from the 2D acquisition technique.

Another thirteen volumes (GM: n = 1, WM: n = 12), including all V_Traj for Patient 9, showed increased R₁Gd (Fig 3B). Notably, for Patient 9, R₁ was lower than otherwise found in WM, no PpIX fluorescence was found, and the neuropathological analysis of the two biopsy samples showed an increase of astrocytes but no sign of malignancy. Other studies have found small relative increases after Gd in non-tumorous tissue, e.g., in contralateral control tissue [24], offering a potential explanation for the observed R₁Gd increase in Patient 9. Another potential influence could be movement during image acquisition, especially along the z-direction.

Notably, R₁ for Patient 7 was comparable to that of NAWM rather than GTV, as indicated by the radiological VOI analysis , and a slight decrease in R₁Gd was observed (Fig 3B). The restricted response could be due to preoperative bleeding, or the majority of necrotic tissue (60%) identified in the neuropathological re-evaluation.

These results support the finding that R₁Gd increase is not tumor-specific and could benefit from combined analysis with other measures.

Increased R₁Gd rates overlap fluorescence peaks and tumor percentage

Over 80% of the PpIX peaks corresponded to an increase in R₁Gd. Interestingly, the PpIX response seems delayed compared to the increase in R₁Gd, see Patients 2, 7, and 8 in Figs 3B and 3C. This could be explained by the volume differences between fluorescence penetration depth (<1mm) and the slice thickness of the qMRI sequence (4 mm). Similar to the R₁Gd increase, PpIX peaks were found in PTE (n = 2) and WM (n = 1) beyond the radiological definition of GTV. PpIX accumulates intracellularly in neoplastic tissue in proximity to a leaky blood-brain barrier. The mechanisms behind tumor selectivity for PpIX accumulation are not fully known but have been attributed to several cellular factors, including altered pH and enzyme concentrations [28].

Tumor tissue heterogeneity is particularly evident in Patients 1, 5, and 7, where the biopsy sample included viable tumor, necrotic, and non-tumor tissue. These heterogeneities result in altered relaxation rates and fluorescence responses compared to if only viable tumor tissue were present. However, to diagnose a glioblastoma, necrosis or microvascular proliferant vessels need to be present in the sample, or since 2021, certain molecular criteria are also accepted [18].

As Novikov et al argue [29], MRI and neuropathology have limited comparative value since the techniques are not sensitive to the same parameters. However, as neuropathology is the current diagnostic tool for tumors, the comparison is not only needed but necessary for exploring the potential of qMRI in clinical applications.

Relaxation rates and proton densities across tissue types

In the radiological VOIs, decreased R₁ in GTV and PTE compared to NAWM were identified, analogous to the prolonged T₁ times previously presented [8]. T₁ lengthens (i.e., R₁ decreases) as the amount of free water increases, e.g., in PTE compared to WM [3] or in necrotic tissue [7]. R₁ also decreases with intracellular water content and increases with the concentration of macromolecules and iron [30,31]. The relaxation rates for NAWM in this study are comparable to those reported in the literature [7,32]. Likewise, the decreased R₂ in GTV was in line with previous findings [8]. R₂ is receptive to dipole interactions with the surrounding tissue [3], decreases with the fraction of free water, and increases with iron content [8].

In normal-appearing brain tissue, Hagiwara and colleagues found a lower PD in WM compared to GM [32], which is corroborated by the present study. Proton density percentages >100 were found in manual radiological VOIs (GTV, PTE), most likely due to their edematous nature. As PD is inversely dependent on the macromolecule concentration, an increase in PD indicates increased free water content in the voxel [31]. Altered PDGd percentages were observed in the GTV VOI, V_Traj GTV, and seven biopsy volumes. As no PD percentage change between acquisitions is expected, this could be an effect of partial volumes or the PD calculations. As PD is calculated based on parameters from the other maps, including flip and saturation angle, repetition and echo times, T₁ and T₂ times, and an intensity scaling factor [32], any uncertainties in the previous measurements will be amplified in PD.

Postoperative imaging

In the present study, the registration errors reported in the navigation system were smaller than the calculated Euclidean distances between the pre- and postoperatively defined trajectories. Although the Euclidean distance is a compound measure, including the registration error during surgery, movement during equipment fastening, potential brain shift, and errors from image co-registration, the discrepancy between the pre- and postoperative trajectory is prominent. As postoperative image is not a part of the clinical routine for needle biopsy patients, and correlation studies often performed on preoperative data, we encourage caution. Especially when imaging at higher spatial resolution, postoperative imaging, and a robust image co-registration pipeline are necessities for accurate interpretation.

Limitations

A 2D sequence with a slice thickness of 4 mm was used for qMRI acquisition in the present study, introducing inevitable partial volume effects (PVE). These effects are further accentuated if the patient has moved along the z-axis. More specifically, major (n = 3), minor (n = 3), and negligible movement (n = 3) were observed in this cohort. The largest translational displacement between qMRI acquisitions was found for Patient 9. In future studies, the use of a 3D sequence is highly encouraged to reduce PVE. In the context of the 2D sequence, attempts were made to minimize the influence of PVE by performing the analysis in the native R₁, R₂, R₁Gd, and R₂Gd spaces. Consequently, no voxel-to-voxel differences were calculated, but rather the distribution changes over each volume (biopsy, V_Traj, and VOIs) were analyzed.

It should be noted that the radiological VOIs (GTV, PTE, and NAWM) contain a large number of voxels (2x10³ - 64x10³). Thus, the statistical differences in qMRI values were not reported for these volumes. The number of measurements in the V_Traj radiological classes was also unbalanced, varying from 8 (GM) to 30 (GTV) measurements. Another limitation was the small size and heterogeneity of the patient cohort, requiring further data collection to confirm the findings of the present study.

Herein, measurements on several different scales are compared, ranging from micrometer sections in the neuropathological analysis to several millimeters for the surgical procedure. The results were reported on the millimeter scale, as this is the scale of significance during surgery.

Conclusions

A methodology for multimodal pre-, intra-, and postoperative data combining radiology, PpIX-fluorescence, and neuropathology on the millimeter scale was extended to qMRI relaxometry. The combined analysis suggests increased R₁Gd and fluorescence peaks as a marker for high-grade tumor tissue, such as glioblastoma and lymphoma, confirmed by neuropathology. Relaxometry could offer further insights into tissue characteristics beyond contrast enhancement on conventional MRI, however, its implications should be interpreted in conjunction with other modalities, e.g., PpIX-fluorescence and neuropathology, until its mechanisms are fully elucidated.

Supporting information

S1 Fig. Detailed image processing pipeline of preoperative clinical and quantitative MRI as well as postoperative MRI or CT imaging.

T₁wGd space is used as reference space unless otherwise stated. ANTs: advanced normalization tools, FOV: field of view, FSL: FMRIB’s Software Library, GD: gadolinium, GM: gray matter, GTV: gross tumor volume, NAWM: normal appearing white matter, PTE: peritumoral edema, syn: synthetic, VOI: volume of interest, w: weighted.

https://doi.org/10.1371/journal.pone.0326765.s001

(DOCX)

S2 Fig. Longitudinal relaxation rate (R₁) and fluorescence ratio in V_Traj along the biopsy trajectories.

Violin plots represent R₁ distribution before (left) and after (right) gadolinium administration, crosses depict the fluorescence ratio, circles signify conventional radiological classification, and gray blocks specify biopsy sampling volume. GM: gray matter, GTV: gross tumor volume, Pat: Patient, PTE: peritumoral edema, WM: white matter.

https://doi.org/10.1371/journal.pone.0326765.s002

(DOCX)

S3 Fig. Transverse relaxation rate (R₂) and fluorescence ratio in VTraj along the biopsy trajectories.

Violin plots represent R₂ distribution before (left) and after (right) gadolinium administration, crosses depict the fluorescence ratio, circles signify conventional radiological classification, and gray blocks specify biopsy sampling volume. GM: gray matter, GTV: gross tumor volume, Pat: Patient, PTE: peritumoral edema, WM: white matter.

https://doi.org/10.1371/journal.pone.0326765.s003

(DOCX)

Acknowledgments

The authors want to acknowledge the clinical staff at the Departments of Neurosurgery, Radiology, and Clinical Pathology at Linköping University Hospital for their support during the study.

References

1. 1. Jackson RJ, Fuller GN, Abi-Said D, Lang FF, Gokaslan ZL, Shi WM, et al. Limitations of stereotactic biopsy in the initial management of gliomas. Neuro Oncol. 2001;3(3):193–200. pmid:11465400

* View Article

* PubMed/NCBI

* Google Scholar

2. 2. Nilsson M, Englund E, Szczepankiewicz F, van Westen D, Sundgren PC. Imaging brain tumour microstructure. Neuroimage. 2018;182:232–50. pmid:29751058

* View Article

* PubMed/NCBI

* Google Scholar

3. 3. Cercignani M, Dowell NG, Tofts PS. Quantitative MRI of the Brain: Principles of Physical Measurement: CRC Press; 2018.

4. 4. Weiskopf N, Edwards LJ, Helms G, Mohammadi S, Kirilina E. Quantitative magnetic resonance imaging of brain anatomy and in vivo histology. Nature Reviews Physics. 2021;3(8):570–88.

* View Article

* Google Scholar

5. 5. Warntjes JBM, Leinhard OD, West J, Lundberg P. Rapid magnetic resonance quantification on the brain: Optimization for clinical usage. Magn Reson Med. 2008;60(2):320–9. pmid:18666127

* View Article

* PubMed/NCBI

* Google Scholar

6. 6. Deoni SC. Quantitative relaxometry of the brain. Top Magn Reson Imaging. 2010;21(2):101–13.

* View Article

* Google Scholar

7. 7. Blystad I, Warntjes JBM, Smedby Ö, Lundberg P, Larsson E-M, Tisell A. Quantitative MRI using relaxometry in malignant gliomas detects contrast enhancement in peritumoral oedema. Sci Rep. 2020;10(1):17986. pmid:33093605

* View Article

* PubMed/NCBI

* Google Scholar

8. 8. Maurer GD, Tichy J, Harter PN, Nöth U, Weise L, Quick-Weller J. Matching Quantitative MRI Parameters with Histological Features of Treatment-Naïve IDH Wild-Type Glioma. Cancers (Basel). 2021;13(16).

* View Article

* Google Scholar

9. 9. Catapano G, Sgulò FG, Seneca V, Iorio G, de Notaris M, di Nuzzo G. Fluorescein-assisted stereotactic needle biopsy of brain tumors: a single-center experience and systematic review. Neurosurg Rev. 2019;42(2):309–18. pmid:29383601

* View Article

* PubMed/NCBI

* Google Scholar

10. 10. Widhalm G, Minchev G, Woehrer A, Preusser M, Kiesel B, Furtner J, et al. Strong 5-aminolevulinic acid-induced fluorescence is a novel intraoperative marker for representative tissue samples in stereotactic brain tumor biopsies. Neurosurg Rev. 2012;35(3):381–91; discussion 391. pmid:22407140

* View Article

* PubMed/NCBI

* Google Scholar

11. 11. Kiesel B, Millesi M, Woehrer A, Furtner J, Bavand A, Roetzer T, et al. 5-ALA-induced fluorescence as a marker for diagnostic tissue in stereotactic biopsies of intracranial lymphomas: experience in 41 patients. Neurosurg Focus. 2018;44(6):E7. pmid:29852770

* View Article

* PubMed/NCBI

* Google Scholar

12. 12. Hadjipanayis CG, Stummer W. 5-ALA and FDA approval for glioma surgery. J Neurooncol. 2019;141(3):479–86. pmid:30644008

* View Article

* PubMed/NCBI

* Google Scholar

13. 13. Haj-Hosseini N, Richter JCO, Milos P, Hallbeck M, Wårdell K. 5-ALA fluorescence and laser Doppler flowmetry for guidance in a stereotactic brain tumor biopsy. Biomed Opt Express. 2018;9(5):2284–96.

* View Article

* Google Scholar

14. 14. Klint E, Richter J, Wårdell K. Combined use of frameless neuronavigation and in situ optical guidance in brain tumor needle biopsies. Brain Sciences. 2023;13(5):809.

* View Article

* Google Scholar

15. 15. Klint E, Richter J, Milos P, Hallbeck M, Wårdell K. In situ optical feedback in brain tumor biopsy: A multiparametric analysis. Neurooncol Adv. 2024;6(1):vdae175. pmid:39687792

* View Article

* PubMed/NCBI

* Google Scholar

16. 16. Fedorov A, Beichel R, Kalpathy-Cramer J, Finet J, Fillion-Robin J-C, Pujol S, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging. 2012;30(9):1323–41. pmid:22770690

* View Article

* PubMed/NCBI

* Google Scholar

17. 17. Klint E, Mauritzon S, Ragnemalm B, Richter J, Wardell K. FluoRa - a System for Combined Fluorescence and Microcirculation Measurements in Brain Tumor Surgery. Annu Int Conf IEEE Eng Med Biol Soc. 2021;2021:1512–5. pmid:34891572

* View Article

* PubMed/NCBI

* Google Scholar

18. 18. WHO Classification of Tumours Editorial Board. Central nervous system tumours. 5 ed. Lyon (France): International Agency for Research on Cancer; 2021.

19. 19. Gorgolewski K, Burns C, Madison C, Clark D, Halchenko Y, Waskom M. Nipype: A Flexible, Lightweight and Extensible Neuroimaging Data Processing Framework in Python. Frontiers in Neuroinformatics. 2011;5.

* View Article

* Google Scholar

20. 20. Jenkinson M, Beckmann CF, Behrens TE, Woolrich MW, Smith SM. FSL. Neuroimage. 2012;62(2):782–90.

* View Article

* Google Scholar

21. 21. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage. 2011;54(3):2033–44. pmid:20851191

* View Article

* PubMed/NCBI

* Google Scholar

22. 22. Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological). 1995;57(1):289–300.

* View Article

* Google Scholar

23. 23. Hattingen E, Müller A, Jurcoane A, Mädler B, Ditter P, Schild H, et al. Value of quantitative magnetic resonance imaging T1-relaxometry in predicting contrast-enhancement in glioblastoma patients. Oncotarget. 2017;8(32):53542–51. pmid:28881830

* View Article

* PubMed/NCBI

* Google Scholar

24. 24. Müller A, Jurcoane A, Kebir S, Ditter P, Schrader F, Herrlinger U, et al. Quantitative T1-mapping detects cloudy-enhancing tumor compartments predicting outcome of patients with glioblastoma. Cancer Med. 2017;6(1):89–99. pmid:27891815

* View Article

* PubMed/NCBI

* Google Scholar

25. 25. Wahsner J, Gale EM, Rodríguez-Rodríguez A, Caravan P. Chemistry of MRI Contrast Agents: Current Challenges and New Frontiers. Chem Rev. 2019;119(2):957–1057. pmid:30350585

* View Article

* PubMed/NCBI

* Google Scholar

26. 26. Rohrer M, Bauer H, Mintorovitch J, Requardt M, Weinmann H-J. Comparison of magnetic properties of MRI contrast media solutions at different magnetic field strengths. Invest Radiol. 2005;40(11):715–24. pmid:16230904

* View Article

* PubMed/NCBI

* Google Scholar

27. 27. Ibrahim MAHB, Dublin AB. Gadolinium Magnetic Resonance Imaging. Treasure Island (FL): StatPearls Publishing; 2023.

28. 28. Collaud S, Juzeniene A, Moan J, Lange N. On the selectivity of 5-aminolevulinic acid-induced protoporphyrin IX formation. Curr Med Chem Anticancer Agents. 2004;4(3):301–16. pmid:15134506

* View Article

* PubMed/NCBI

* Google Scholar

29. 29. Novikov DS, Kiselev VG, Jespersen SN. On modeling. Magn Reson Med. 2018;79(6):3172–93. pmid:29493816

* View Article

* PubMed/NCBI

* Google Scholar

30. 30. Rooney WD, Johnson G, Li X, Cohen ER, Kim SG, Ugurbil K. Magnetic field and tissue dependencies of human brain longitudinal 1H2O relaxation in vivo. Magnetic Resonance in Medicine. 2007;57(2):308–18.

* View Article

* Google Scholar

31. 31. Reuter G, Lommers E, Balteau E, Simon J, Phillips C, Scholtes F, et al. Multiparameter quantitative histological MRI values in high-grade gliomas: a potential biomarker of tumor progression. Neurooncol Pract. 2020;7(6):646–55. pmid:33304600

* View Article

* PubMed/NCBI

* Google Scholar

32. 32. Hagiwara A, Hori M, Cohen-Adad J, Nakazawa M, Suzuki Y, Kasahara A, et al. Linearity, Bias, Intrascanner Repeatability, and Interscanner Reproducibility of Quantitative Multidynamic Multiecho Sequence for Rapid Simultaneous Relaxometry at 3 T: A Validation Study With a Standardized Phantom and Healthy Controls. Invest Radiol. 2019;54(1):39–47. pmid:30300164

* View Article

* PubMed/NCBI

* Google Scholar

Citation: Klint E, Tisell A, Blystad I, Hallbeck M, Nordin T, Hillman J, et al. (2025) Quantitative MRI relaxometry in brain tumor needle biopsies: Multimodal comparison with tissue fluorescence, radiology, and neuropathology. PLoS One 20(7): e0326765. https://doi.org/10.1371/journal.pone.0326765

About the Authors:

Elisabeth Klint

Roles: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

E-mail: [email protected]

Affiliation: Department of Biomedical Engineering, Linköping University, Sweden

ORICD: https://orcid.org/0000-0002-7987-9841

Anders Tisell

Roles: Data curation, Investigation, Methodology, Writing – review & editing

Affiliations: Department of Radiation Physics in Linköping, and Department of Health, Medicine and Caring Sciences, Linköping University, Sweden, Centre for Medical Image Science and Visualization (CMIV), Linköping University, Sweden

Ida Blystad

Roles: Data curation, Formal analysis, Methodology, Writing – review & editing

Affiliations: Centre for Medical Image Science and Visualization (CMIV), Linköping University, Sweden, Department of Radiology in Linköping and Department of Health, Medicine and Caring Sciences, Linköping University, Sweden

ORICD: https://orcid.org/0000-0002-8857-5698

Martin Hallbeck

Roles: Data curation, Formal analysis, Methodology, Writing – review & editing

Affiliation: Department of Clinical Pathology in Linköping, and Department of Biomedical and Clinical Sciences, Linköping University, Sweden

Teresa Nordin

Roles: Formal analysis, Methodology, Supervision, Writing – review & editing

Affiliation: Department of Biomedical Engineering, Linköping University, Sweden

ORICD: https://orcid.org/0000-0002-1641-9848

Jan Hillman

Roles: Conceptualization, Funding acquisition, Writing – review & editing

Affiliation: Department of Neurosurgery in Linköping, and Department of Biomedical and Clinical Sciences, Linköping University, Sweden

Johan Richter

Roles: Data curation, Investigation, Methodology, Supervision, Validation, Writing – review & editing

Affiliation: Department of Neurosurgery in Linköping, and Department of Biomedical Engineering, Linköping University, Sweden

ORICD: https://orcid.org/0000-0001-6222-0875

Karin Wårdell

Roles: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing

Affiliation: Department of Biomedical Engineering, Linköping University, Sweden

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

References

1. Jackson RJ, Fuller GN, Abi-Said D, Lang FF, Gokaslan ZL, Shi WM, et al. Limitations of stereotactic biopsy in the initial management of gliomas. Neuro Oncol. 2001;3(3):193–200. pmid:11465400

2. Nilsson M, Englund E, Szczepankiewicz F, van Westen D, Sundgren PC. Imaging brain tumour microstructure. Neuroimage. 2018;182:232–50. pmid:29751058

3. Cercignani M, Dowell NG, Tofts PS. Quantitative MRI of the Brain: Principles of Physical Measurement: CRC Press; 2018.

4. Weiskopf N, Edwards LJ, Helms G, Mohammadi S, Kirilina E. Quantitative magnetic resonance imaging of brain anatomy and in vivo histology. Nature Reviews Physics. 2021;3(8):570–88.

5. Warntjes JBM, Leinhard OD, West J, Lundberg P. Rapid magnetic resonance quantification on the brain: Optimization for clinical usage. Magn Reson Med. 2008;60(2):320–9. pmid:18666127

6. Deoni SC. Quantitative relaxometry of the brain. Top Magn Reson Imaging. 2010;21(2):101–13.

7. Blystad I, Warntjes JBM, Smedby Ö, Lundberg P, Larsson E-M, Tisell A. Quantitative MRI using relaxometry in malignant gliomas detects contrast enhancement in peritumoral oedema. Sci Rep. 2020;10(1):17986. pmid:33093605

8. Maurer GD, Tichy J, Harter PN, Nöth U, Weise L, Quick-Weller J. Matching Quantitative MRI Parameters with Histological Features of Treatment-Naïve IDH Wild-Type Glioma. Cancers (Basel). 2021;13(16).

9. Catapano G, Sgulò FG, Seneca V, Iorio G, de Notaris M, di Nuzzo G. Fluorescein-assisted stereotactic needle biopsy of brain tumors: a single-center experience and systematic review. Neurosurg Rev. 2019;42(2):309–18. pmid:29383601

10. Widhalm G, Minchev G, Woehrer A, Preusser M, Kiesel B, Furtner J, et al. Strong 5-aminolevulinic acid-induced fluorescence is a novel intraoperative marker for representative tissue samples in stereotactic brain tumor biopsies. Neurosurg Rev. 2012;35(3):381–91; discussion 391. pmid:22407140

11. Kiesel B, Millesi M, Woehrer A, Furtner J, Bavand A, Roetzer T, et al. 5-ALA-induced fluorescence as a marker for diagnostic tissue in stereotactic biopsies of intracranial lymphomas: experience in 41 patients. Neurosurg Focus. 2018;44(6):E7. pmid:29852770

12. Hadjipanayis CG, Stummer W. 5-ALA and FDA approval for glioma surgery. J Neurooncol. 2019;141(3):479–86. pmid:30644008

13. Haj-Hosseini N, Richter JCO, Milos P, Hallbeck M, Wårdell K. 5-ALA fluorescence and laser Doppler flowmetry for guidance in a stereotactic brain tumor biopsy. Biomed Opt Express. 2018;9(5):2284–96.

14. Klint E, Richter J, Wårdell K. Combined use of frameless neuronavigation and in situ optical guidance in brain tumor needle biopsies. Brain Sciences. 2023;13(5):809.

15. Klint E, Richter J, Milos P, Hallbeck M, Wårdell K. In situ optical feedback in brain tumor biopsy: A multiparametric analysis. Neurooncol Adv. 2024;6(1):vdae175. pmid:39687792

16. Fedorov A, Beichel R, Kalpathy-Cramer J, Finet J, Fillion-Robin J-C, Pujol S, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging. 2012;30(9):1323–41. pmid:22770690

17. Klint E, Mauritzon S, Ragnemalm B, Richter J, Wardell K. FluoRa - a System for Combined Fluorescence and Microcirculation Measurements in Brain Tumor Surgery. Annu Int Conf IEEE Eng Med Biol Soc. 2021;2021:1512–5. pmid:34891572

18. WHO Classification of Tumours Editorial Board. Central nervous system tumours. 5 ed. Lyon (France): International Agency for Research on Cancer; 2021.

19. Gorgolewski K, Burns C, Madison C, Clark D, Halchenko Y, Waskom M. Nipype: A Flexible, Lightweight and Extensible Neuroimaging Data Processing Framework in Python. Frontiers in Neuroinformatics. 2011;5.

20. Jenkinson M, Beckmann CF, Behrens TE, Woolrich MW, Smith SM. FSL. Neuroimage. 2012;62(2):782–90.

21. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage. 2011;54(3):2033–44. pmid:20851191

22. Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological). 1995;57(1):289–300.

23. Hattingen E, Müller A, Jurcoane A, Mädler B, Ditter P, Schild H, et al. Value of quantitative magnetic resonance imaging T1-relaxometry in predicting contrast-enhancement in glioblastoma patients. Oncotarget. 2017;8(32):53542–51. pmid:28881830

24. Müller A, Jurcoane A, Kebir S, Ditter P, Schrader F, Herrlinger U, et al. Quantitative T1-mapping detects cloudy-enhancing tumor compartments predicting outcome of patients with glioblastoma. Cancer Med. 2017;6(1):89–99. pmid:27891815

25. Wahsner J, Gale EM, Rodríguez-Rodríguez A, Caravan P. Chemistry of MRI Contrast Agents: Current Challenges and New Frontiers. Chem Rev. 2019;119(2):957–1057. pmid:30350585

26. Rohrer M, Bauer H, Mintorovitch J, Requardt M, Weinmann H-J. Comparison of magnetic properties of MRI contrast media solutions at different magnetic field strengths. Invest Radiol. 2005;40(11):715–24. pmid:16230904

27. Ibrahim MAHB, Dublin AB. Gadolinium Magnetic Resonance Imaging. Treasure Island (FL): StatPearls Publishing; 2023.

28. Collaud S, Juzeniene A, Moan J, Lange N. On the selectivity of 5-aminolevulinic acid-induced protoporphyrin IX formation. Curr Med Chem Anticancer Agents. 2004;4(3):301–16. pmid:15134506

29. Novikov DS, Kiselev VG, Jespersen SN. On modeling. Magn Reson Med. 2018;79(6):3172–93. pmid:29493816

30. Rooney WD, Johnson G, Li X, Cohen ER, Kim SG, Ugurbil K. Magnetic field and tissue dependencies of human brain longitudinal 1H2O relaxation in vivo. Magnetic Resonance in Medicine. 2007;57(2):308–18.

31. Reuter G, Lommers E, Balteau E, Simon J, Phillips C, Scholtes F, et al. Multiparameter quantitative histological MRI values in high-grade gliomas: a potential biomarker of tumor progression. Neurooncol Pract. 2020;7(6):646–55. pmid:33304600

32. Hagiwara A, Hori M, Cohen-Adad J, Nakazawa M, Suzuki Y, Kasahara A, et al. Linearity, Bias, Intrascanner Repeatability, and Interscanner Reproducibility of Quantitative Multidynamic Multiecho Sequence for Rapid Simultaneous Relaxometry at 3 T: A Validation Study With a Standardized Phantom and Healthy Controls. Invest Radiol. 2019;54(1):39–47. pmid:30300164