Morphometric analysis of white matter lesions in MR images: method and validation

GNN-based structural information to improve DNN-based basal ganglia segmentation in children following early brain lesion

Analyzing the basal ganglia following an early brain lesion is crucial due to their noteworthy role in sensory-motor functions. However, the segmentation of these subcortical structures on MRI is challenging in children and is further complicated by the presence of a lesion. Although current deep neural networks (DNN) perform well in segmenting subcortical brain structures in healthy brains, they lack robustness when faced with lesion variability, leading to structural inconsistencies. Given the established spatial organization of the basal ganglia, we propose enhancing the DNN-based segmentation through post-processing with a graph neural network (GNN). The GNN conducts node classification on graphs encoding both class probabilities and spatial information regarding the regions segmented by the DNN. In this study, we focus on neonatal arterial ischemic stroke (NAIS) in children. The approach is evaluated on both healthy children and children after NAIS using three DNN backbones: U-Net, UNETr, and MSGSE-Net. The results show an improvement in segmentation performance, with an increase in the median Dice score by up to 4% and a reduction in the median Hausdorff distance (HD) by up to 93% for healthy children (from 36.45 to 2.57) and up to 91% for children suffering from NAIS (from 40.64 to 3.50). The performance of the method is compared with atlas-based methods. Severe cases of neonatal stroke result in a decline in performance in the injured hemisphere, without negatively affecting the segmentation of the contra-injured hemisphere. Furthermore, the approach demonstrates resilience to small training datasets, a widespread challenge in the medical field, particularly in pediatrics and for rare pathologies.

Evaluating tooth segmentation accuracy and time efficiency in CBCT images using artificial intelligence: A systematic review and Meta-analysis

OBJECTIVES

This systematic review and meta-analysis aimed to assess the current performance of artificial intelligence (AI)-based methods for tooth segmentation in three-dimensional cone-beam computed tomography (CBCT) images, with a focus on their accuracy and efficiency compared to those of manual segmentation techniques.

DATA

The data analyzed in this review consisted of a wide range of research studies utilizing AI algorithms for tooth segmentation in CBCT images. Meta-analysis was performed, focusing on the evaluation of the segmentation results using the dice similarity coefficient (DSC).

SOURCES

PubMed, Embase, Scopus, Web of Science, and IEEE Explore were comprehensively searched to identify relevant studies. The initial search yielded 5642 entries, and subsequent screening and selection processes led to the inclusion of 35 studies in the systematic review. Among the various segmentation methods employed, convolutional neural networks, particularly the U-net model, are the most commonly utilized. The pooled effect of the DSC score for tooth segmentation was 0.95 (95 %CI 0.94 to 0.96). Furthermore, seven papers provided insights into the time required for segmentation, which ranged from 1.5 s to 3.4 min when utilizing AI techniques.

CONCLUSIONS

AI models demonstrated favorable accuracy in automatically segmenting teeth from CBCT images while reducing the time required for the process. Nevertheless, correction methods for metal artifacts and tooth structure segmentation using different imaging modalities should be addressed in future studies.

CLINICAL SIGNIFICANCE

AI algorithms have great potential for precise tooth measurements, orthodontic treatment planning, dental implant placement, and other dental procedures that require accurate tooth delineation. These advances have contributed to improved clinical outcomes and patient care in dental practice.

A novel evaluation model of image registration for cone-beam computed tomography guided lung cancer radiotherapy

BACKGROUND

The aim of the study was to establish a weighted comprehensive evaluation model (WCEM) of image registration for cone-beam computed tomography (CBCT) guided lung cancer radiotherapy that considers the geometric accuracy of gross target volume (GTV) and organs at risk (OARs), and assess the registration accuracy of different image registration methods to provide clinical references.

METHODS

The planning CT and CBCT images of 20 lung cancer patients were registered using diverse algorithms (bony and grayscale) and regions of interest (target, ipsilateral, and body). We compared the coverage ratio (CR) of the planning target volume (PTVCT) to GTVCBCT, as well as the dice similarity coefficient (DSC) of the GTV and OARs, considering the treatment position across various registration methods. Furthermore, we developed a mathematical model to assess registration results comprehensively. This model was evaluated and validated using CRFs across four automatic registration methods.

RESULTS

The grayscale registration method, coupled with the registration of the ipsilateral structure, exhibited the highest level of automatic registration accuracy, the DSC were 0.87 ± 0.09 (GTV), 0.71 ± 0.09 (esophagus), 0.74 ± 0.09 (spinal cord), and 0.91 ± 0.05 (heart), respectively. Our proposed WCEM proved to be both practical and effective. The results clearly indicated that the grayscale registration method, when applied to the ipsilateral structure, achieved the highest CRF score. The average CRF scores, excellent rates, good rate and qualification rates were 58 ± 26, 40%, 75%, and 85%, respectively.

CONCLUSIONS

This study successfully developed a clinically relevant weighted evaluation model for CBCT-guided lung cancer radiotherapy. Validation confirmed the grayscale method's optimal performance in ipsilateral structure registration.

Development and validation of a reliable method for automated measurements of psoas muscle volume in CT scans using deep learning-based segmentation: a cross-sectional study

OBJECTIVES

We aimed to develop an automated method for measuring the volume of the psoas muscle using CT to aid sarcopenia research efficiently.

METHODS

We used a data set comprising the CT scans of 520 participants who underwent health check-ups at a health promotion centre. We developed a psoas muscle segmentation model using deep learning in a three-step process based on the nnU-Net method. The automated segmentation method was evaluated for accuracy, reliability, and time required for the measurement.

RESULTS

The Dice similarity coefficient was used to compare the manual segmentation with automated segmentation; an average Dice score of 0.927 ± 0.019 was obtained, with no critical outliers. Our automated segmentation system had an average measurement time of 2 min 20 s ± 20 s, which was 48 times shorter than that of the manual measurement method (111 min 6 s ± 25 min 25 s).

CONCLUSION

We have successfully developed an automated segmentation method to measure the psoas muscle volume that ensures consistent and unbiased estimates across a wide range of CT images.

When Two Eyes Don't Suffice-Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

Hyperfluorescence (HF) and reduced autofluorescence (RA) are important biomarkers in fundus autofluorescence images (FAF) for the assessment of health of the retinal pigment epithelium (RPE), an important indicator of disease progression in geographic atrophy (GA) or central serous chorioretinopathy (CSCR). Autofluorescence images have been annotated by human raters, but distinguishing biomarkers (whether signals are increased or decreased) from the normal background proves challenging, with borders being particularly open to interpretation. Consequently, significant variations emerge among different graders, and even within the same grader during repeated annotations. Tests on in-house FAF data show that even highly skilled medical experts, despite previously discussing and settling on precise annotation guidelines, reach a pair-wise agreement measured in a Dice score of no more than 63-80% for HF segmentations and only 14-52% for RA. The data further show that the agreement of our primary annotation expert with herself is a 72% Dice score for HF and 51% for RA. Given these numbers, the task of automated HF and RA segmentation cannot simply be refined to the improvement in a segmentation score. Instead, we propose the use of a segmentation ensemble. Learning from images with a single annotation, the ensemble reaches expert-like performance with an agreement of a 64-81% Dice score for HF and 21-41% for RA with all our experts. In addition, utilizing the mean predictions of the ensemble networks and their variance, we devise ternary segmentations where FAF image areas are labeled either as confident background, confident HF, or potential HF, ensuring that predictions are reliable where they are confident (97% Precision), while detecting all instances of HF (99% Recall) annotated by all experts.

The role of 18F-FDG PET in minimizing variability in gross tumor volume delineation of soft tissue sarcomas

BACKGROUND

Accurate gross tumor volume (GTV) delineation is a critical step in radiation therapy treatment planning. However, it is reader dependent and thus susceptible to intra- and inter-reader variability. GTV delineation of soft tissue sarcoma (STS) often relies on CT and MR images.

PURPOSE

This study investigates the potential role of 18F-FDG PET in reducing intra- and inter-reader variability thereby improving reproducibility of GTV delineation in STS, without incurring additional costs or radiation exposure.

MATERIALS AND METHODS

Three readers performed independent GTV delineation of 61 patients with STS using first CT and MR followed by CT, MR, and 18F-FDG PET images. Each reader performed a total of six delineation trials, three trials per imaging modality group. Dice Similarity Coefficient (DSC) score and Hausdorff distance (HD) were used to assess both intra- and inter-reader variability using generated simultaneous truth and performance level estimation (STAPLE) GTVs as ground truth. Statistical analysis was performed using a Wilcoxon signed-ranked test.

RESULTS

There was a statistically significant decrease in both intra- and inter-reader variability in GTV delineation using CT, MR 18F-FDG PET images vs. CT and MR images. This was translated by an increase in the DSC score and a decrease in the HD for GTVs drawn from CT, MR and 18F-FDG PET images vs. GTVs drawn from CT and MR for all readers and across all three trials.

CONCLUSION

Incorporation of 18F-FDG PET into CT and MR images decreased intra- and inter-reader variability and subsequently increased reproducibility of GTV delineation in STS.

The brain markers of creativity measured by divergent thinking in childhood: Hippocampal volume and functional connectivity

Creativity, a high-order cognitive ability, has received wide attention from researchers and educators who are dedicated to promoting its development throughout one's lifespan. Currently, creativity is commonly assessed with divergent thinking tasks, such as the Alternative Uses Task. Recent advancements in neuroimaging techniques have enabled the identification of brain markers for high-order cognitive abilities. One such brain structure of interest in this regard is the hippocampus, which has been found to play an important role in generating creative thoughts in adulthood. However, such role of the hippocampus in childhood is not clear. Thus, this study aimed to investigate the associations between creativity, as measured by divergent thinking, and both the volume of the hippocampus and its resting-state functional connectivity in 116 children aged 8-12 years. The results indicate significant relations between divergent thinking and the volume of the hippocampal head and the hippocampal tail, as well as the volume of a subfield comprising cornu ammonis 2-4 and dentate gyrus within the hippocampal body. Additionally, divergent thinking was significantly related to the differences between the anterior and the posterior hippocampus in their functional connectivity to other brain regions during rest. These results suggest that these two subregions may collaborate with different brain regions to support diverse cognitive processes involved in the generation of creative thoughts. In summary, these findings indicate that divergent thinking is significantly related to the structural and functional characteristics of the hippocampus, offering potential insights into the brain markers for creativity during the developmental stage.

Spatial correspondence of cortical activity measured with whole head fNIRS and fMRI: Toward clinical use within subject

Functional near infrared spectroscopy (fNIRS) and functional magnetic resonance imaging (fMRI) both measure the hemodynamic response, and so both imaging modalities are expected to have a strong correspondence in regions of cortex adjacent to the scalp. To assess whether fNIRS can be used clinically in a manner similar to fMRI, 22 healthy adult participants underwent same-day fMRI and whole-head fNIRS testing while they performed separate motor (finger tapping) and visual (flashing checkerboard) tasks. Analyses were conducted within and across subjects for each imaging approach, and regions of significant task-related activity were compared on the cortical surface. The spatial correspondence between fNIRS and fMRI detection of task-related activity was good in terms of true positive rate, with fNIRS overlap of up to 68 % of the fMRI for analyses across subjects (group analysis) and an average overlap of up to 47.25 % for individual analyses within subject. At the group level, the positive predictive value of fNIRS was 51 % relative to fMRI. The positive predictive value for within subject analyses was lower (41.5 %), reflecting the presence of significant fNIRS activity in regions without significant fMRI activity. This could reflect task-correlated sources of physiologic noise and/or differences in the sensitivity of fNIRS and fMRI measures to changes in separate (vs. combined) measures of oxy and de-oxyhemoglobin. The results suggest whole-head fNIRS as a noninvasive imaging modality with promising clinical utility for the functional assessment of brain activity in superficial regions of cortex physically adjacent to the skull.

Structural Connectivity of the Human Piriform Cortex: an Exploratory Study

BACKGROUND AND OBJECTIVES

The piriform cortex (PC) is part of the primary olfactory network in humans. Recent findings suggest that it plays a role in pathophysiology of epilepsy. Therefore, studying its connectivity can further our understanding of seizure propagation in epilepsy. We aimed to explore the structural connectivity of PC using high-quality human connectome project data coupled with segmentation of PC on anatomic MRI.

METHODS

Twenty subjects were randomly selected from the human connectome project database, and PC was traced on each hemisphere. Probabilistic whole-brain tractography was then used to visualize PC connectivity.

RESULTS

The strongest connectivity was noted between PC and ipsilateral insula in both hemispheres. Specifically, the posterior long gyrus of each insula was predominantly connected to PC. This was followed by connections between PC and basal ganglia as well as orbital frontal cortices.

CONCLUSION

The PC has the strongest connectivity with the insula bilaterally. Specifically, the posterior long gyri of insula have the strongest connectivity. This finding may provide additional insight for localizing and treating temporo-insular epilepsy.

Shape-constrained deformable brain segmentation: Methods and quantitative validation

MRI-guided neuro interventions require rapid, accurate, and reproducible segmentation of anatomical brain structures for identification of targets during surgical procedures and post-surgical evaluation of intervention efficiency. Segmentation algorithms must be validated and cleared for clinical use. This work introduces a methodology for shape-constrained deformable brain segmentation, describes the quantitative validation used for its clinical clearance, and presents a comparison with manual expert segmentation and FreeSurfer, an open source software for neuroimaging data analysis. ClearPoint Maestro is software for fully-automatic brain segmentation from T1-weighted MRI that combines a shape-constrained deformable brain model with voxel-wise tissue segmentation within the cerebral hemispheres and the cerebellum. The performance of the segmentation was validated in terms of accuracy and reproducibility. Segmentation accuracy was evaluated with respect to training data and independently traced ground truth. Segmentation reproducibility was quantified and compared with manual expert segmentation and FreeSurfer. Quantitative reproducibility analysis indicates superior performance compared to both manual expert segmentation and FreeSurfer. The shape-constrained methodology results in accurate and highly reproducible segmentation. Inherent point based-correspondence provides consistent target identification ideal for MRI-guided neuro interventions.

Deep Learning-Assisted Diffusion Tensor Imaging for Evaluation of the Physis and Metaphysis

Diffusion tensor imaging of physis and metaphysis can be used as a biomarker to predict height change in the pediatric population. Current application of this technique requires manual segmentation of the physis which is time-consuming and introduces interobserver variability. UNET Transformers (UNETR) can be used for automatic segmentation to optimize workflow. Three hundred and eighty-five DTI scans from 191 subjects with mean age of 12.6 years ± 2.01 years were retrospectively used for training and validation. The mean Dice correlation coefficient was 0.81 for the UNETR model and 0.68 for the UNET. Manual extraction and segmentation took 15 min per volume, whereas both deep learning segmentation techniques took < 1 s per volume and were deterministic, always producing the same result for a given input. Intraclass correlation coefficient (ICC) for ROI-derived femur diffusion metrics was excellent for tract count (0.95), volume (0.95), and FA (0.97), and good for tract length (0.87). The results support the hypothesis that a hybrid UNETR model can be trained to replace the manual segmentation of physeal DTI images, therefore automating the process.

Optic chiasm manual and automated measurements in sub-acute optic neuritis with OCT and MRI correlations

PURPOSE

The optic chiasm (OC) is a central structure in the visual pathway and can be visualized in conventional MRI, but no consensus regarding its measurement has been defined. We aim to investigate the most reproducible manual approach to OC measurement and to explore associations of OC with optical coherence tomography (OCT) parameters, and automatic brain segmentation (FreeSurfer) in subacute optic neuritis (sON), multiple sclerosis without optic neuritis (MSwoON), and healthy subjects (HS).

MATERIALS AND METHODS

We reproduced two previously reported methodologies and implemented a new proposed simplified approach, entitled optic chiasm mean area (OCMA). The intra and inter-rater reliability and reproducibility were assessed through the intraclass correlation (ICC) and Dice similarity (DSC) coefficients. Partial correlations were calculated to gauge the associations between OCMA fraction (OCMA divided by total intracranial volume), brain regional segmentations derived from FreeSurfer, and OCT parameters.

RESULTS

We have analysed 43 sON, 20 MSwoON, and 20 HS. OCMA presented better results for reliability in both intra- and inter-rater analysis (excellent ICC and DSC with over 80% overlap between masks), as compared to the other two approaches. OCMA fraction was associated with OC volume fraction obtained with Freesurfer in all groups, brain parenchymal fraction, and OCT parameters in MSwoON.

CONCLUSIONS

The OCMA is a simplified approach to measure OC atrophy, has a higher reliability than the current approaches and shows association with an automated method. OC-derived measures seem to reflect diffuse neurodegenerative damage, whereas, in patients with subacute ON, it may be associated with local damage.

Expert Agreement on the Presence and Spatial Localization of Melanocytic Features in Dermoscopy

Dermoscopy aids in melanoma detection; however, agreement on dermoscopic features, including those of high clinical relevance, remains poor. In this study, we attempted to evaluate agreement among experts on exemplar images not only for the presence of melanocytic-specific features but also for spatial localization. This was a cross-sectional, multicenter, observational study. Dermoscopy images exhibiting at least 1 of 31 melanocytic-specific features were submitted by 25 world experts as exemplars. Using a web-based platform that allows for image markup of specific contrast-defined regions (superpixels), 20 expert readers annotated 248 dermoscopic images in collections of 62 images. Each collection was reviewed by five independent readers. A total of 4,507 feature observations were performed. Good-to-excellent agreement was found for 14 of 31 features (45.2%), with eight achieving excellent agreement (Gwet's AC >0.75) and seven of them being melanoma-specific features. These features were peppering/granularity (0.91), shiny white streaks (0.89), typical pigment network (0.83), blotch irregular (0.82), negative network (0.81), irregular globules (0.78), dotted vessels (0.77), and blue-whitish veil (0.76). By utilizing an exemplar dataset, a good-to-excellent agreement was found for 14 features that have previously been shown useful in discriminating nevi from melanoma. All images are public (www.isic-archive.com) and can be used for education, scientific communication, and machine learning experiments.

Image harmonization improves consistency of intra-rater delineations of MS lesions in heterogeneous MRI

Clinical magnetic resonance images (MRIs) lack a standard intensity scale due to differences in scanner hardware and the pulse sequences used to acquire the images. When MRIs are used for quantification, as in the evaluation of white matter lesions (WMLs) in multiple sclerosis, this lack of intensity standardization becomes a critical problem affecting both the staging and tracking of the disease and its treatment. This paper presents a study of harmonization on WML segmentation consistency, which is evaluated using an object detection classification scheme that incorporates manual delineations from both the original and harmonized MRIs. A cohort of ten people scanned on two different imaging platforms was studied. An expert rater, blinded to the image source, manually delineated WMLs on images from both scanners before and after harmonization. It was found that there is closer agreement in both global and per-lesion WML volume and spatial distribution after harmonization, demonstrating the importance of image harmonization prior to the creation of manual delineations. These results could lead to better truth models in both the development and evaluation of automated lesion segmentation algorithms.

Two-dimensional segmentation fusion tool: an extensible, free-to-use, user-friendly tool for combining different bidimensional segmentations

Introduction: In several fields, the process of fusing multiple two-dimensional (2D) closed lines is an important step. For instance, this is fundamental in histology and oncology in general. The treatment of a tumor consists of numerous steps and activities. Among them, segmenting the cancer area, that is, the correct identification of its spatial location by the segmentation technique, is one of the most important and at the same time complex and delicate steps. The difficulty in deriving reliable segmentations stems from the lack of a standard for identifying the edges and surrounding tissues of the tumor area. For this reason, the entire process is affected by considerable subjectivity. Given a tumor image, different practitioners can associate different segmentations with it, and the diagnoses produced may differ. Moreover, experimental data show that the analysis of the same area by the same physician at two separate timepoints may result in different lines being produced. Accordingly, it is challenging to establish which contour line is the ground truth. Methods: Starting from multiple segmentations related to the same tumor, statistical metrics and computational procedures could be exploited to combine them for determining the most reliable contour line. In particular, numerous algorithms have been developed over time for this procedure, but none of them is validated yet. Accordingly, in this field, there is no ground truth, and research is still active. Results: In this work, we developed the Two-Dimensional Segmentation Fusion Tool (TDSFT), a user-friendly tool distributed as a free-to-use standalone application for MAC, Linux, and Windows, which offers a simple and extensible interface where numerous algorithms are proposed to "compute the mean" (i.e., the process to fuse, combine, and "average") multiple 2D lines. Conclusions: The TDSFT can support medical specialists, but it can also be used in other fields where it is required to combine 2D close lines. In addition, the TDSFT is designed to be easily extended with new algorithms thanks to a dedicated graphical interface for configuring new parameters. The TDSFT can be downloaded from the following link: https://sourceforge.net/p/tdsft.

Automatic tooth periodontal ligament segmentation of cone beam computed tomography based on instance segmentation network

Objective

The three-dimensional morphological structures of periodontal ligaments (PDLs) are important data for periodontal, orthodontic, prosthodontic, and implant interventions. This study aimed to employ a deep learning (DL) algorithm to segment the PDL automatically in cone-beam computed tomography (CBCT).

Method

This was a retrospective study. We randomly selected 389 patients and 1734 axial CBCT images from the CBCT database, and designed a fully automatic PDL segmentation computer-aided model based on instance segmentation Mask R-CNN network. The labels of the model training were 'teeth' and 'alveolar bone', and the 'PDL' is defined as the region where the 'teeth' and 'alveolar bone' overlap. The model's segmentation performance was evaluated using CBCT data from eight patients outside the database.

Results

Qualitative evaluation indicates that the PDL segmentation accuracy of incisors, canines, premolars, wisdom teeth, and implants reached 100%. The segmentation accuracy of molars was 96.4%. Quantitative evaluation indicates that the mIoU and mDSC of PDL segmentation were 0.667 ± 0.015 (>0.6) and 0.799 ± 0.015 (>0.7) respectively.

Conclusion

This study analysed a unique approach to AI-driven automatic segmentation of PDLs on CBCT imaging, possibly enabling chair-side measurements of PDLs to facilitate periodontists, orthodontists, prosthodontists, and implantologists in more efficient and accurate diagnosis and treatment planning.

A Perspective on Developing Modeling and Image Analysis Tools to Investigate Mechanosensing Proteins

The shift of funding organizations to prioritize interdisciplinary work points to the need for workflow models that better accommodate interdisciplinary studies. Most scientists are trained in a specific field and are often unaware of the kind of insights that other disciplines could contribute to solving various problems. In this paper, we present a perspective on how we developed an experimental pipeline between a microscopy and image analysis/bioengineering lab. Specifically, we connected microscopy observations about a putative mechanosensing protein, obscurin, to image analysis techniques that quantify cell changes. While the individual methods used are well established (fluorescence microscopy; ImageJ WEKA and mTrack2 programs; MATLAB), there are no existing best practices for how to integrate these techniques into a cohesive, interdisciplinary narrative. Here, we describe a broadly applicable workflow of how microscopists can more easily quantify cell properties (e.g., perimeter, velocity) from microscopy videos of eukaryotic (MDCK) adherent cells. Additionally, we give examples of how these foundational measurements can create more complex, customizable cell mechanics tools and models.

Crowdsourcing Skin Demarcations of Chronic Graft-Versus-Host Disease in Patient Photographs: Training Versus Performance Study

BACKGROUND

Chronic graft-versus-host disease (cGVHD) is a significant cause of long-term morbidity and mortality in patients after allogeneic hematopoietic cell transplantation. Skin is the most commonly affected organ, and visual assessment of cGVHD can have low reliability. Crowdsourcing data from nonexpert participants has been used for numerous medical applications, including image labeling and segmentation tasks.

OBJECTIVE

This study aimed to assess the ability of crowds of nonexpert raters-individuals without any prior training for identifying or marking cGHVD-to demarcate photos of cGVHD-affected skin. We also studied the effect of training and feedback on crowd performance.

METHODS

Using a Canfield Vectra H1 3D camera, 360 photographs of the skin of 36 patients with cGVHD were taken. Ground truth demarcations were provided in 3D by a trained expert and reviewed by a board-certified dermatologist. In total, 3000 2D images (projections from various angles) were created for crowd demarcation through the DiagnosUs mobile app. Raters were split into high and low feedback groups. The performances of 4 different crowds of nonexperts were analyzed, including 17 raters per image for the low and high feedback groups, 32-35 raters per image for the low feedback group, and the top 5 performers for each image from the low feedback group.

RESULTS

Across 8 demarcation competitions, 130 raters were recruited to the high feedback group and 161 to the low feedback group. This resulted in a total of 54,887 individual demarcations from the high feedback group and 78,967 from the low feedback group. The nonexpert crowds achieved good overall performance for segmenting cGVHD-affected skin with minimal training, achieving a median surface area error of less than 12% of skin pixels for all crowds in both the high and low feedback groups. The low feedback crowds performed slightly poorer than the high feedback crowd, even when a larger crowd was used. Tracking the 5 most reliable raters from the low feedback group for each image recovered a performance similar to that of the high feedback crowd. Higher variability between raters for a given image was not found to correlate with lower performance of the crowd consensus demarcation and cannot therefore be used as a measure of reliability. No significant learning was observed during the task as more photos and feedback were seen.

CONCLUSIONS

Crowds of nonexpert raters can demarcate cGVHD images with good overall performance. Tracking the top 5 most reliable raters provided optimal results, obtaining the best performance with the lowest number of expert demarcations required for adequate training. However, the agreement amongst individual nonexperts does not help predict whether the crowd has provided an accurate result. Future work should explore the performance of crowdsourcing in standard clinical photos and further methods to estimate the reliability of consensus demarcations.

Analysis of preoperative computed tomography radiomics and clinical factors for predicting postsurgical recurrence of papillary thyroid carcinoma

BACKGROUND

Postsurgical recurrence is of great concern for papillary thyroid carcinoma (PTC). We aim to investigate the value of computed tomography (CT)-based radiomics features and conventional clinical factors in predicting the recurrence of PTC.

METHODS

Two-hundred and eighty patients with PTC were retrospectively enrolled and divided into training and validation cohorts at a 6:4 ratio. Recurrence was defined as cytology/pathology-proven disease or morphological evidence of lesions on imaging examinations within 5 years after surgery. Radiomics features were extracted from manually segmented tumor on CT images and were then selected using four different feature selection methods sequentially. Multivariate logistic regression analysis was conducted to identify clinical features associated with recurrence. Radiomics, clinical, and combined models were constructed separately using logistic regression (LR), support vector machine (SVM), k-nearest neighbor (KNN), and neural network (NN), respectively. Receiver operating characteristic analysis was performed to evaluate the model performance in predicting recurrence. A nomogram was established based on all relevant features, with its reliability and reproducibility verified using calibration curves and decision curve analysis (DCA).

RESULTS

Eighty-nine patients with PTC experienced recurrence. A total of 1218 radiomics features were extracted from each segmentation. Five radiomics and six clinical features were related to recurrence. Among the 4 radiomics models, the LR-based and SVM-based radiomics models outperformed the NN-based radiomics model (P = 0.032 and 0.026, respectively). Among the 4 clinical models, only the difference between the area under the curve (AUC) of the LR-based and NN-based clinical model was statistically significant (P = 0.035). The combined models had higher AUCs than the corresponding radiomics and clinical models based on the same classifier, although most differences were not statistically significant. In the validation cohort, the combined models based on the LR, SVM, KNN, and NN classifiers had AUCs of 0.746, 0.754, 0.669, and 0.711, respectively. However, the AUCs of these combined models had no significant differences (all P > 0.05). Calibration curves and DCA indicated that the nomogram have potential clinical utility.

CONCLUSIONS

The combined model may have potential for better prediction of PTC recurrence than radiomics and clinical models alone. Further testing with larger cohort may help reach statistical significance.

A (Sub)field Guide to Quality Control in Hippocampal Subfield Segmentation on Highresolution T2-weighted MRI

Inquiries into properties of brain structure and function have progressed due to developments in magnetic resonance imaging (MRI). To sustain progress in investigating and quantifying neuroanatomical details in vivo, the reliability and validity of brain measurements are paramount. Quality control (QC) is a set of procedures for mitigating errors and ensuring the validity and reliability of brain measurements. Despite its importance, there is little guidance on best QC practices and reporting procedures. The study of hippocampal subfields in vivo is a critical case for QC because of their small size, inter-dependent boundary definitions, and common artifacts in the MRI data used for subfield measurements. We addressed this gap by surveying the broader scientific community studying hippocampal subfields on their views and approaches to QC. We received responses from 37 investigators spanning 10 countries, covering different career stages, and studying both healthy and pathological development and aging. In this sample, 81% of researchers considered QC to be very important or important, and 19% viewed it as fairly important. Despite this, only 46% of researchers reported on their QC processes in prior publications. In many instances, lack of reporting appeared due to ambiguous guidance on relevant details and guidance for reporting, rather than absence of QC. Here, we provide recommendations for correcting errors to maximize reliability and minimize bias. We also summarize threats to segmentation accuracy, review common QC methods, and make recommendations for best practices and reporting in publications. Implementing the recommended QC practices will collectively improve inferences to the larger population, as well as have implications for clinical practice and public health.

Deep learning models for automatic tumor segmentation and total tumor volume assessment in patients with colorectal liver metastases

BACKGROUND

We developed models for tumor segmentation to automate the assessment of total tumor volume (TTV) in patients with colorectal liver metastases (CRLM).

METHODS

In this prospective cohort study, pre- and post-systemic treatment computed tomography (CT) scans of 259 patients with initially unresectable CRLM of the CAIRO5 trial (NCT02162563) were included. In total, 595 CT scans comprising 8,959 CRLM were divided into training (73%), validation (6.5%), and test sets (21%). Deep learning models were trained with ground truth segmentations of the liver and CRLM. TTV was calculated based on the CRLM segmentations. An external validation cohort was included, comprising 72 preoperative CT scans of patients with 112 resectable CRLM. Image segmentation evaluation metrics and intraclass correlation coefficient (ICC) were calculated.

RESULTS

In the test set (122 CT scans), the autosegmentation models showed a global Dice similarity coefficient (DSC) of 0.96 (liver) and 0.86 (CRLM). The corresponding median per-case DSC was 0.96 (interquartile range [IQR] 0.95-0.96) and 0.80 (IQR 0.67-0.87). For tumor segmentation, the intersection-over-union, precision, and recall were 0.75, 0.89, and 0.84, respectively. An excellent agreement was observed between the reference and automatically computed TTV for the test set (ICC 0.98) and external validation cohort (ICC 0.98). In the external validation, the global DSC was 0.82 and the median per-case DSC was 0.60 (IQR 0.29-0.76) for tumor segmentation.

CONCLUSIONS

Deep learning autosegmentation models were able to segment the liver and CRLM automatically and accurately in patients with initially unresectable CRLM, enabling automatic TTV assessment in such patients.

RELEVANCE STATEMENT

Automatic segmentation enables the assessment of total tumor volume in patients with colorectal liver metastases, with a high potential of decreasing radiologist's workload and increasing accuracy and consistency.

KEY POINTS

• Tumor response evaluation is time-consuming, manually performed, and ignores total tumor volume. • Automatic models can accurately segment tumors in patients with colorectal liver metastases. • Total tumor volume can be accurately calculated based on automatic segmentations.

Subject-Specific Automatic Reconstruction of White Matter Tracts

MRI-based tractography is still underexploited and unsuited for routine use in brain tumor surgery due to heterogeneity of methods and functional-anatomical definitions and above all, the lack of a turn-key system. Standardization of methods is therefore desirable, whereby an objective and reliable approach is a prerequisite before the results of any automated procedure can subsequently be validated and used in neurosurgical practice. In this work, we evaluated these preliminary but necessary steps in healthy volunteers. Specifically, we evaluated the robustness and reliability (i.e., test-retest reproducibility) of tractography results of six clinically relevant white matter tracts by using healthy volunteer data (N = 136) from the Human Connectome Project consortium. A deep learning convolutional network-based approach was used for individualized segmentation of regions of interest, combined with an evidence-based tractography protocol and appropriate post-tractography filtering. Robustness was evaluated by estimating the consistency of tractography probability maps, i.e., averaged tractograms in normalized space, through the use of a hold-out cross-validation approach. No major outliers were found, indicating a high robustness of the tractography results. Reliability was evaluated at the individual level. First by examining the overlap of tractograms that resulted from repeatedly processed identical MRI scans (N = 10, 10 iterations) to establish an upper limit of reliability of the pipeline. Second, by examining the overlap for subjects that were scanned twice at different time points (N = 40). Both analyses indicated high reliability, with the second analysis showing a reliability near the upper limit. The robust and reliable subject-specific generation of white matter tracts in healthy subjects holds promise for future validation of our pipeline in a clinical population and subsequent implementation in brain tumor surgery.

Deep Learning Artificial Intelligence Tool for Automated Radiographic Determination of Posterior Tibial Slope in Patients With ACL Injury

Background

An increased posterior tibial slope (PTS) corresponds with an increased risk of graft failure after anterior cruciate ligament (ACL) reconstruction (ACLR). Validated methods of manual PTS measurements are subject to potential interobserver variability and can be inefficient on large datasets.

Purpose/Hypothesis

To develop a deep learning artificial intelligence technique for automated PTS measurement from standard lateral knee radiographs. It was hypothesized that this deep learning tool would be able to measure the PTS on a high volume of radiographs expeditiously and that these measurements would be similar to previously validated manual measurements.

Study Design

Cohort study (diagnosis); Level of evidence, 2.

Methods

A deep learning U-Net model was developed on a cohort of 300 postoperative short-leg lateral radiographs from patients who underwent ACLR to segment the tibial shaft, tibial joint surface, and tibial tuberosity. The model was trained via a random split after an 80 to 20 train-validation scheme. Masks for training images were manually segmented, and the model was trained for 400 epochs. An image processing pipeline was then deployed to annotate and measure the PTS using the predicted segmentation masks. Finally, the performance of this combined pipeline was compared with human measurements performed by 2 study personnel using a previously validated manual technique for measuring the PTS on short-leg lateral radiographs on an independent test set consisting of both pre- and postoperative images.

Results

The U-Net semantic segmentation model achieved a mean Dice similarity coefficient of 0.885 on the validation cohort. The mean difference between the human-made and computer-vision measurements was 1.92° (σ = 2.81° [P = .24]). Extreme disagreements between the human and machine measurements, as defined by ≥5° differences, occurred <5% of the time. The model was incorporated into a web-based digital application front-end for demonstration purposes, which can measure a single uploaded image in Portable Network Graphics format in a mean time of 5 seconds.

Conclusion

We developed an efficient and reliable deep learning computer vision algorithm to automate the PTS measurement on short-leg lateral knee radiographs. This tool, which demonstrated good agreement with human annotations, represents an effective clinical adjunct for measuring the PTS as part of the preoperative assessment of patients with ACL injuries.

Application of FGD-BCEL loss function in segmenting temporal lobes on localized CT images for radiotherapy

Objectives

The aim of this study was to find a new loss function to automatically segment temporal lobes on localized CT images for radiotherapy with more accuracy and a solution to dealing with the classification of class-imbalanced samples in temporal lobe segmentation.

Methods

Localized CT images for radiotherapy of 70 patients with nasopharyngeal carcinoma were selected. Radiation oncologists sketched mask maps. The dataset was randomly divided into the training set (n = 49), the validation set (n = 7), and the test set (n = 14). The training set was expanded by rotation, flipping, zooming, and shearing, and the models were evaluated using Dice similarity coefficient (DSC), Jaccard similarity coefficient (JSC), positive predictive value (PPV), sensitivity (SE), and Hausdorff distance (HD). This study presented an improved loss function, focal generalized Dice-binary cross-entropy loss (FGD-BCEL), and compared it with four other loss functions, Dice loss (DL), generalized Dice loss (GDL), Tversky loss (TL), and focal Tversky loss (FTL), using the U-Net model framework.

Results

With the U-Net model based on FGD-BCEL, the DSC, JSC, PPV, SE, and HD were 0.87 ± 0.11, 0.78 ± 0.11, 0.90 ± 0.10, 0.87 ± 0.13, and 4.11 ± 0.75, respectively. Except for the SE, all the other evaluation metric values of the temporal lobes segmented by the FGD-BCEL-based U-Net model were improved compared to the DL, GDL, TL, and FTL loss function-based U-Net models. Moreover, the FGD-BCEL-based U-Net model was morphologically more similar to the mask maps. The over- and under-segmentation was lessened, and it effectively segmented the tiny structures in the upper and lower poles of the temporal lobe with a limited number of samples.

Conclusions

For the segmentation of the temporal lobe on localized CT images for radiotherapy, the U-Net model based on the FGD-BCEL can meet the basic clinical requirements and effectively reduce the over- and under-segmentation compared with the U-Net models based on the other four loss functions. However, there still exists some over- and under-segmentation in the results, and further improvement is needed.

Numerical approach to investigate MR imaging artifacts from orthopedic implants at different field strengths according to ASTM F2119

OBJECTIVE

This study presents an extended evaluation of a numerical approach to simulate artifacts of metallic implants in an MR environment.

METHODS

The numerical approach is validated by comparing the artifact shape of the simulations and measurements of two metallic orthopedic implants at three different field strengths (1.5 T, 3 T, and 7 T). Furthermore, this study presents three additional use cases of the numerical simulation. The first one shows how numerical simulations can improve the artifact size evaluation according to ASTM F2119. The second use case quantifies the influence of different imaging parameters (TE and bandwidth) on the artifact size. Finally, the third use case shows the potential of performing human model artifact simulations.

RESULTS

The numerical simulation approach shows a dice similarity coefficient of 0.74 between simulated and measured artifact sizes of metallic implants. The alternative artifact size calculation method presented in this study shows that the artifact size of the ASTM-based method is up to 50% smaller for complex shaped implants compared to the numerical-based approach.

CONCLUSION

In conclusion, the numerical approach could be used in the future to extend MR safety testing according to a revision of the ASTM F2119 standard and for design optimization during the development process of implants.

Development and validation of a quality control procedure for automatic segmentation of hippocampal subfields

Automatic segmentation methods for in vivo magnetic resonance imaging are increasing in popularity because of their high efficiency and reproducibility. However, automatic methods can be perfectly reliable and consistently wrong, and the validity of automatic segmentation methods cannot be taken for granted. Quality control (QC) by trained and reliable human raters is necessary to ensure the validity of automatic measurements. Yet QC practices for applied neuroimaging research are underdeveloped. We report a detailed QC and correction procedure to accompany our validated atlas for hippocampal subfield segmentation. We document a two-step QC procedure for identifying segmentation errors, along with a taxonomy of errors and an error severity rating scale. This detailed procedure has high between-rater reliability for error identification and manual correction. The latter introduces at maximum 3% error variance in volume measurement. All procedures were cross-validated on an independent sample collected at a second site with different imaging parameters. The analysis of error frequency revealed no evidence of bias. An independent rater with a third sample replicated procedures with high within-rater reliability for error identification and correction. We provide recommendations for implementing the described method along with hypothesis testing strategies. In sum, we present a detailed QC procedure that is optimized for efficiency while prioritizing measurement validity and suits any automatic atlas.

An active learning approach to train a deep learning algorithm for tumor segmentation from brain MR images

PURPOSE

This study focuses on assessing the performance of active learning techniques to train a brain MRI glioma segmentation model.

METHODS

The publicly available training dataset provided for the 2021 RSNA-ASNR-MICCAI Brain Tumor Segmentation (BraTS) Challenge was used in this study, consisting of 1251 multi-institutional, multi-parametric MR images. Post-contrast T1, T2, and T2 FLAIR images as well as ground truth manual segmentation were used as input for the model. The data were split into a training set of 1151 cases and testing set of 100 cases, with the testing set remaining constant throughout. Deep convolutional neural network segmentation models were trained using the NiftyNet platform. To test the viability of active learning in training a segmentation model, an initial reference model was trained using all 1151 training cases followed by two additional models using only 575 cases and 100 cases. The resulting predicted segmentations of these two additional models on the remaining training cases were then addended to the training dataset for additional training.

RESULTS

It was demonstrated that an active learning approach for manual segmentation can lead to comparable model performance for segmentation of brain gliomas (0.906 reference Dice score vs 0.868 active learning Dice score) while only requiring manual annotation for 28.6% of the data.

CONCLUSION

The active learning approach when applied to model training can drastically reduce the time and labor spent on preparation of ground truth training data.

CRITICAL RELEVANCE STATEMENT

Active learning concepts were applied to a deep learning-assisted segmentation of brain gliomas from MR images to assess their viability in reducing the required amount of manually annotated ground truth data in model training.

KEY POINTS

• This study focuses on assessing the performance of active learning techniques to train a brain MRI glioma segmentation model. • The active learning approach for manual segmentation can lead to comparable model performance for segmentation of brain gliomas. • Active learning when applied to model training can drastically reduce the time and labor spent on preparation of ground truth training data.

Altered structural connectivity in olfactory disfunction after mild COVID-19 using probabilistic tractography

We aimed to investigate changes in olfactory bulb volume and brain network in the white matter (WM) in patients with persistent olfactory disfunction (OD) following COVID-19. A cross-sectional study evaluated 38 participants with OD after mild COVID-19 and 24 controls, including Sniffin' Sticks identification test (SS-16), MoCA, and brain magnetic resonance imaging. Network-Based Statistics (NBS) and graph theoretical analysis were used to explore the WM. The COVID-19 group had reduced olfactory bulb volume compared to controls. In NBS, COVID-19 patients showed increased structural connectivity in a subnetwork comprising parietal brain regions. Regarding global network topological properties, patients exhibited lower global and local efficiency and higher assortativity than controls. Concerning local network topological properties, patients had reduced local efficiency (left lateral orbital gyrus and pallidum), increased clustering (left lateral orbital gyrus), increased nodal strength (right anterior orbital gyrus), and reduced nodal strength (left amygdala). SS-16 test score was negatively correlated with clustering of whole-brain WM in the COVID-19 group. Thus, patients with OD after COVID-19 had relevant WM network dysfunction with increased connectivity in the parietal sensory cortex. Reduced integration and increased segregation are observed within olfactory-related brain areas might be due to compensatory plasticity mechanisms devoted to recovering olfactory function.

Deep learning image segmentation approaches for malignant bone lesions: a systematic review and meta-analysis

Introduction

Image segmentation is an important process for quantifying characteristics of malignant bone lesions, but this task is challenging and laborious for radiologists. Deep learning has shown promise in automating image segmentation in radiology, including for malignant bone lesions. The purpose of this review is to investigate deep learning-based image segmentation methods for malignant bone lesions on Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Positron-Emission Tomography/CT (PET/CT).

Method

The literature search of deep learning-based image segmentation of malignant bony lesions on CT and MRI was conducted in PubMed, Embase, Web of Science, and Scopus electronic databases following the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). A total of 41 original articles published between February 2017 and March 2023 were included in the review.

Results

The majority of papers studied MRI, followed by CT, PET/CT, and PET/MRI. There was relatively even distribution of papers studying primary vs. secondary malignancies, as well as utilizing 3-dimensional vs. 2-dimensional data. Many papers utilize custom built models as a modification or variation of U-Net. The most common metric for evaluation was the dice similarity coefficient (DSC). Most models achieved a DSC above 0.6, with medians for all imaging modalities between 0.85-0.9.

Discussion

Deep learning methods show promising ability to segment malignant osseous lesions on CT, MRI, and PET/CT. Some strategies which are commonly applied to help improve performance include data augmentation, utilization of large public datasets, preprocessing including denoising and cropping, and U-Net architecture modification. Future directions include overcoming dataset and annotation homogeneity and generalizing for clinical applicability.

Preoperative Radiotherapy with a Simultaneous Integrated Boost Compared to Chemoradiotherapy for cT3-4 Rectal Cancer: Long-Term Results of a Multicenter Randomized Study

BACKGROUND

Preoperative chemoradiotherapy (CRT) is the standard treatment for T3-4 rectal cancer. Here, we compared image-guided and intensity-modulated RT (IG-IMRT) with a simultaneous integrated boost (SIB) (instead of concomitant chemotherapy) versus CRT in a multi-centric randomized trial.

METHODS

cT3-4 rectal cancer patients were randomly assigned to receive preoperative IG-IMRT 46 Gy/23 fractions plus capecitabine 825 mg/m² twice daily (CRT arm) or IG-IMRT 46 Gy/23 fractions with an SIB to the rectal tumor up to a total dose of 55.2 Gy (RTSIB arm).

RESULTS

A total of 174 patients were randomly assigned between April 2010 and May 2014. Grade 3 acute toxicities were 6% and 4% in the CRT and RTSIB arms, respectively. The mean fractional change in SUVmax at 5 weeks after completion of preoperative RT were -55.8% (±24.0%) and -52.9% (±21.6%) for patients in the CRT arm and RTSIB arm, respectively (p = 0.43). The pathologic complete response rate was 24% with CRT compared to 14% with RTSIB. There were no differences in 5-year overall survival (OS), progression-free survival (PFS) or local control (LC).

CONCLUSIONS

The preoperative RTSIB approach was not inferior to CRT in terms of metabolic response, toxicity, OS, PFS and LC, and could be considered an available option for patients unfit for fluorouracil-based CRT.

Leveraging 2D Deep Learning ImageNet-trained models for Native 3D Medical Image Analysis

Convolutional neural networks (CNNs) have shown promising performance in various 2D computer vision tasks due to availability of large amounts of 2D training data. Contrarily, medical imaging deals with 3D data and usually lacks the equivalent extent and diversity of data, for developing AI models. Transfer learning provides the means to use models trained for one application as a starting point to another application. In this work, we leverage 2D pre-trained models as a starting point in 3D medical applications by exploring the concept of Axial-Coronal-Sagittal (ACS) convolutions. We have incorporated ACS as an alternative of native 3D convolutions in the Generally Nuanced Deep Learning Framework (GaNDLF), providing various well-established and state-of-the-art network architectures with the availability of pre-trained encoders from 2D data. Results of our experimental evaluation on 3D MRI data of brain tumor patients for i) tumor segmentation and ii) radiogenomic classification, show model size reduction by ~22% and improvement in validation accuracy by ~33%. Our findings support the advantage of ACS convolutions in pre-trained 2D CNNs over 3D CNN without pre-training, for 3D segmentation and classification tasks, democratizing existing models trained in datasets of unprecedented size and showing promise in the field of healthcare.

Comparative analysis of automatic segmentation of esophageal cancer using 3D Res-UNet on conventional and 40-keV virtual mono-energetic CT Images: a retrospective study

Objectives

To assess the performance of 3D Res-UNet for fully automated segmentation of esophageal cancer (EC) and compare the segmentation accuracy between conventional images (CI) and 40-keV virtual mono-energetic images (VMI_40 kev).

Methods

Patients underwent spectral CT scanning and diagnosed of EC by operation or gastroscope biopsy in our hospital from 2019 to 2020 were analyzed retrospectively. All artery spectral base images were transferred to the dedicated workstation to generate VMI_40 kev and CI. The segmentation model of EC was constructed by 3D Res-UNet neural network in VMI_40 kev and CI, respectively. After optimization training, the Dice similarity coefficient (DSC), overlap (IOU), average symmetrical surface distance (ASSD) and 95% Hausdorff distance (HD_95) of EC at pixel level were tested and calculated in the test set. The paired rank sum test was used to compare the results of VMI_40 kev and CI.

Results

A total of 160 patients were included in the analysis and randomly divided into the training dataset (104 patients), validation dataset (26 patients) and test dataset (30 patients). VMI_40 kevas input data in the training dataset resulted in higher model performance in the test dataset in comparison with using CI as input data (DSC:0.875 vs 0.859, IOU: 0.777 vs 0.755, ASSD:0.911 vs 0.981, HD_95: 4.41 vs 6.23, all p-value <0.05).

Conclusion

Fully automated segmentation of EC with 3D Res-UNet has high accuracy and clinically feasibility for both CI and VMI_40 kev. Compared with CI, VMI_40 kev indicated slightly higher accuracy in this test dataset.

Ultrasound image segmentation using Gamma combined with Bayesian model for focused-ultrasound-surgery lesion recognition

This study aims to investigate the feasibility of combined segmentation for the separation of lesions from non-ablated regions, which allows surgeons to easily distinguish, measure, and evaluate the lesion area, thereby improving the quality of high-intensity focused-ultrasound (HIFU) surgery used for the non-invasive tumor treatment. Given that the flexible shape of the Gamma mixture model (GΓMM) fits the complex statistical distribution of samples, a method combining the GΓMM and Bayes framework is constructed for the classification of samples to obtain the segmentation result. An appropriate normalization range and parameters can be used to rapidly obtain a good performance of GΓMM segmentation. The performance values of the proposed method under four metrics (Dice score: 85%, Jaccard coefficient: 75%, recall: 86%, and accuracy: 96%) are better than those of conventional approaches including Otsu and Region growing. Furthermore, the statistical result of sample intensity indicates that the finding of the GΓMM is similar to that obtained by the manual method. These results indicate the stability and reliability of the GΓMM combined with the Bayes framework for the segmentation of HIFU lesions in ultrasound images. The experimental results show the possibility of combining the GΓMM with the Bayes framework to segment lesion areas and evaluate the effect of therapeutic ultrasound.

A survey on deep learning for skin lesion segmentation

Skin cancer is a major public health problem that could benefit from computer-aided diagnosis to reduce the burden of this common disease. Skin lesion segmentation from images is an important step toward achieving this goal. However, the presence of natural and artificial artifacts (e.g., hair and air bubbles), intrinsic factors (e.g., lesion shape and contrast), and variations in image acquisition conditions make skin lesion segmentation a challenging task. Recently, various researchers have explored the applicability of deep learning models to skin lesion segmentation. In this survey, we cross-examine 177 research papers that deal with deep learning-based segmentation of skin lesions. We analyze these works along several dimensions, including input data (datasets, preprocessing, and synthetic data generation), model design (architecture, modules, and losses), and evaluation aspects (data annotation requirements and segmentation performance). We discuss these dimensions both from the viewpoint of select seminal works, and from a systematic viewpoint, examining how those choices have influenced current trends, and how their limitations should be addressed. To facilitate comparisons, we summarize all examined works in a comprehensive table as well as an interactive table available online³.

Tooth automatic segmentation from CBCT images: a systematic review

OBJECTIVES

To describe the current state of the art regarding technological advances in full-automatic tooth segmentation approaches from 3D cone-beam computed tomography (CBCT) images.

MATERIALS AND METHODS

In March 2023, a search strategy without a timeline setting was carried out through a combination of MeSH terms and free text words pooled through Boolean operators ('AND', 'OR') on the following databases: PubMed, Scopus, Web of Science and IEEE Explore. Randomized and non-randomized controlled trials, cohort, case-control, cross-sectional and retrospective studies in the English language only were included.

RESULTS

The search strategy identified 541 articles, of which 23 have been selected. The most employed segmentation methods were based on deep learning approaches. One article exposed an automatic approach for tooth segmentation based on a watershed algorithm and another article used an improved level set method. Four studies presented classical machine learning and thresholding approaches. The most employed metric for evaluating segmentation performance was the Dice similarity index which ranged from 90 ± 3% to 97.9 ± 1.5%.

CONCLUSIONS

Thresholding appeared not reliable for tooth segmentation from CBCT images, whereas convolutional neural networks (CNNs) have been demonstrated as the most promising approach. CNNs could help overcome tooth segmentation's main limitations from CBCT images related to root anatomy, heavy scattering, immature teeth, metal artifacts and time consumption. New studies with uniform protocols and evaluation metrics with random sampling and blinding for data analysis are encouraged to objectively compare the different deep learning architectures' reliability.

CLINICAL RELEVANCE

Automatic tooth segmentation's best performance has been obtained through CNNs for the different ambits of digital dentistry.

Quantifying numerical and spatial reliability of hippocampal and amygdala subdivisions in FreeSurfer

On-going, large-scale neuroimaging initiatives can aid in uncovering neurobiological causes and correlates of poor mental health, disease pathology, and many other important conditions. As projects grow in scale with hundreds, even thousands, of individual participants and scans collected, quantification of brain structures by automated algorithms is becoming the only truly tractable approach. Here, we assessed the spatial and numerical reliability for newly deployed automated segmentation of hippocampal subfields and amygdala nuclei in FreeSurfer 7. In a sample of participants with repeated structural imaging scans (N = 928), we found numerical reliability (as assessed by intraclass correlations, ICCs) was reasonable. Approximately 95% of hippocampal subfields had "excellent" numerical reliability (ICCs ≥ 0.90), while only 67% of amygdala subnuclei met this same threshold. In terms of spatial reliability, 58% of hippocampal subfields and 44% of amygdala subnuclei had Dice coefficients ≥ 0.70. Notably, multiple regions had poor numerical and/or spatial reliability. We also examined correlations between spatial reliability and person-level factors (e.g., participant age; T1 image quality). Both sex and image scan quality were related to variations in spatial reliability metrics. Examined collectively, our work suggests caution should be exercised for a few hippocampal subfields and amygdala nuclei with more variable reliability.

Application of Automatic Segmentation on Super-Resolution Reconstruction MR Images of the Abnormal Fetal Brain

BACKGROUND AND PURPOSE

Fetal brain MR imaging is clinically used to characterize fetal brain abnormalities. Recently, algorithms have been proposed to reconstruct high-resolution 3D fetal brain volumes from 2D slices. By means of these reconstructions, convolutional neural networks have been developed for automatic image segmentation to avoid labor-intensive manual annotations, usually trained on data of normal fetal brains. Herein, we tested the performance of an algorithm specifically developed for segmentation of abnormal fetal brains.

MATERIALS AND METHODS

This was a single-center retrospective study on MR images of 16 fetuses with severe CNS anomalies (gestation, 21-39 weeks). T2-weighted 2D slices were converted to 3D volumes using a super-resolution reconstruction algorithm. The acquired volumetric data were then processed by a novel convolutional neural network to perform segmentations of white matter and the ventricular system and cerebellum. These were compared with manual segmentation using the Dice coefficient, Hausdorff distance (95th percentile), and volume difference. Using interquartile ranges, we identified outliers of these metrics and further analyzed them in detail.

RESULTS

The mean Dice coefficient was 96.2%, 93.7%, and 94.7% for white matter and the ventricular system and cerebellum, respectively. The Hausdorff distance was 1.1, 2.3, and 1.6 mm, respectively. The volume difference was 1.6, 1.4, and 0.3 mL, respectively. Of the 126 measurements, there were 16 outliers among 5 fetuses, discussed on a case-by-case basis.

CONCLUSIONS

Our novel segmentation algorithm obtained excellent results on MR images of fetuses with severe brain abnormalities. Analysis of the outliers shows the need to include pathologies underrepresented in the current data set. Quality control to prevent occasional errors is still needed.

Does consensus contours improve robustness and accuracy on [Formula: see text]F-FDG PET imaging tumor delineation?

PURPOSE

The aim of this study is to explore the robustness and accuracy of consensus contours with 225 nasopharyngeal carcinoma (NPC) clinical cases and 13 extended cardio-torso simulated lung tumors (XCAT) based on 2-deoxy-2-[[Formula: see text]F]fluoro-D-glucose ([Formula: see text]F-FDG) PET imaging.

METHODS

Primary tumor segmentation was performed with two different initial masks on 225 NPC [Formula: see text]F-FDG PET datasets and 13 XCAT simulations using methods of automatic segmentation with active contour, affinity propagation (AP), contrast-oriented thresholding (ST), and 41% maximum tumor value (41MAX), respectively. Consensus contours (ConSeg) were subsequently generated based on the majority vote rule. The metabolically active tumor volume (MATV), relative volume error (RE), Dice similarity coefficient (DSC) and their respective test-retest (TRT) metrics between different masks were adopted to analyze the results quantitatively. The nonparametric Friedman and post hoc Wilcoxon tests with Bonferroni adjustment for multiple comparisons were performed with [Formula: see text] 0.05 considered to be significant.

RESULTS

AP presented the highest variability for MATV in different masks, and ConSeg presented much better TRT performances in MATV compared with AP, and slightly poorer TRT in MATV compared with ST or 41MAXin most cases. Similar trends were also found in RE and DSC with the simulated data. The average of four segmentation results (AveSeg) showed better or comparable results in accuracy for most cases with respect to ConSeg. AP, AveSeg and ConSeg presented better RE and DSC in irregular masks as compared with rectangle masks. Additionally, all methods underestimated the tumour boundaries in relation to the ground truth for XCAT including respiratory motion.

CONCLUSIONS

The consensus method could be a robust approach to alleviate segmentation variabilities, but did not seem to improve the accuracy of segmentation results on average. Irregular initial masks might be at least in some cases attributable to mitigate the segmentation variability as well.

Index lesion contouring on prostate MRI for targeted MRI/US fusion biopsy - Evaluation of mismatch between radiologists and urologists

PURPOSE

Mistargeting of focal lesions due to inaccurate segmentations can lead to false-negative findings on MRI-guided targeted biopsies. The purpose of this retrospective study was to examine inter-reader agreement of prostate index lesion segmentations from actual biopsy data between urologists and radiologists.

METHOD

Consecutive patients undergoing transperineal MRI-targeted prostate biopsy for PI-RADS 3-5 lesions between January 2020 and December 2021 were included. Agreement between segmentations on T2w-images between urologists and radiologists was assessed with Dice similarity coefficient (DSC) and 95 % Hausdorff distance (95 % HD). Differences in similarity scores were compared using Wilcoxon test. Differences depending on lesion features (size, zonal location, PI-RADS scores, lesion distinctness) were tested with Mann-Whitney U test. Correlation with prostate signal-intensity homogeneity score (PSHS) and lesion size was tested with Spearman's rank correlation.

RESULTS

Ninety-three patients (mean age 64.9 ± 7.1y, median serum PSA 6.5 [4.33-10.00]) were included. Mean similarity scores were statistically significantly lower between urologists and radiologists compared to radiologists only (DSC 0.41 ± 0.24 vs. 0.59 ± 0.23, p < 0.01; 95 %HD 6.38 ± 5.45 mm vs. 4.47 ± 4.12 mm, p < 0.01). There was a moderate and strong positive correlation between DSC scores and lesion size for segmentations from urologists and radiologists (ρ = 0.331, p = 0.002) and radiologists only (ρ = 0.501, p < 0.001). Similarity scores were worse in lesions ≤ 10 mm while other lesion features did not significantly influence similarity scores.

CONCLUSION

There is significant mismatch of prostate index lesion segmentations between urologists and radiologists. Segmentation agreement positively correlates with lesion size. PI-RADS scores, zonal location, lesion distinctness, and PSHS show no significant impact on segmentation agreement. These findings could underpin benefits of perilesional biopsies.

Tensor- and high-resolution fiber tractography for the delineation of the optic radiation and corticospinal tract in the proximity of intracerebral lesions: a reproducibility and repeatability study

PURPOSE

Fiber tracking (FT) is used in neurosurgical planning for the resection of lesions in proximity to fiber pathways, as it contributes to a substantial amelioration of postoperative neurological impairments. Currently, diffusion-tensor imaging (DTI)-based FT is the most frequently used technique; however, sophisticated techniques such as Q-ball (QBI) for high-resolution FT (HRFT) have suggested favorable results. Little is known about the reproducibility of both techniques in the clinical setting. Therefore, this study aimed to examine the intra- and interrater agreement for the depiction of white matter pathways such as the corticospinal tract (CST) and the optic radiation (OR).

METHODS

Nineteen patients with eloquent lesions in the proximity of the OR or CST were prospectively enrolled. Two different raters independently reconstructed the fiber bundles by applying probabilistic DTI- and QBI-FT. Interrater agreement was evaluated from the comparison between results obtained by the two raters on the same data set acquired in two independent iterations at different timepoints using the Dice Similarity Coefficient (DSC) and the Jaccard Coefficient (JC). Likewise, intrarater agreement was determined for each rater comparing individual results.

RESULTS

DSC values showed substantial intrarater agreement based on DTI-FT (rater 1: mean 0.77 (0.68-0.85); rater 2: mean 0.75 (0.64-0.81); p = 0.673); while an excellent agreement was observed after the deployment of QBI-based FT (rater 1: mean 0.86 (0.78-0.98); rater 2: mean 0.80 (0.72-0.91); p = 0.693). In contrast, fair agreement was observed between both measures for the repeatability of the OR of each rater based on DTI-FT (rater 1: mean 0.36 (0.26-0.77); rater 2: mean 0.40 (0.27-0.79), p = 0.546). A substantial agreement between the measures was noted by applying QBI-FT (rater 1: mean 0.67 (0.44-0.78); rater 2: mean 0.62 (0.32-0.70), 0.665). The interrater agreement was moderate for the reproducibility of the CST and OR for both DSC and JC based on DTI-FT (DSC and JC ≥ 0.40); while a substantial interrater agreement was noted for DSC after applying QBI-based FT for the delineation of both fiber tracts (DSC > 0.6).

CONCLUSIONS

Our findings suggest that QBI-based FT might be a more robust tool for the visualization of the OR and CST adjacent to intracerebral lesions compared with the common standard DTI-FT. For neurosurgical planning during the daily workflow, QBI appears to be feasible and less operator-dependent.

Common carotid segmentation in 18 F-sodium fluoride PET/CT scans: Head-to-head comparison of artificial intelligence-based and manual method

BACKGROUND

Carotid atherosclerosis is a major cause of stroke, traditionally diagnosed late. Positron emission tomography/computed tomography (PET/CT) with ¹⁸ F-sodium fluoride (NaF) detects arterial wall micro-calcification long before macro-calcification becomes detectable by ultrasound, CT or magnetic resonance imaging. However, manual PET/CT processing is time-consuming and requires experience. We compared a convolutional neural network (CNN) approach with manual segmentation of the common carotids.

METHODS

Segmentation in NaF-PET/CT scans of 29 healthy volunteers and 20 angina pectoris patients were compared for segmented volume (Vol) and mean, maximal, and total standardized uptake values (SUVmean, SUVmax, and SUVtotal). SUVmean was the average of SUVmeans within the VOI, SUVmax the highest SUV in all voxels in the VOI, and SUVtotal the SUVmean multiplied by the Vol of the VOI. Intra and Interobserver variability with manual segmentation was examined in 25 randomly selected scans.

RESULTS

Bias for Vol, SUVmean, SUVmax, and SUVtotal were 1.33 ± 2.06, -0.01 ± 0.05, 0.09 ± 0.48, and 1.18 ± 1.99 in the left and 1.89 ± 1.5, -0.07 ± 0.12, 0.05 ± 0.47, and 1.61 ± 1.47, respectively, in the right common carotid artery. Manual segmentation lasted typically 20 min versus 1 min with the CNN-based approach. Mean Vol deviation at repeat manual segmentation was 14% and 27% in left and right common carotids.

CONCLUSIONS

CNN-based segmentation was much faster and provided SUVmean values virtually identical to manually obtained ones, suggesting CNN-based analysis as a promising substitute of slow and cumbersome manual processing.

Multi-path decoder U-Net: A weakly trained real-time segmentation network for object detection and localization in ultrasound scans

Detecting and localizing an anatomical structure of interest within the field of view of an ultrasound scan is an essential step in many diagnostic and therapeutic procedures. However, ultrasound scans suffer from high levels of variabilities across sonographers and patients, making it challenging for sonographers to accurately identify and locate these structures without extensive experience. Segmentation-based convolutional neural networks (CNNs) have been proposed as a solution to assist sonographers in this task. Despite their accuracy, these networks require pixel-wise annotations for training; an expensive and labor-intensive operation that requires the expertise of an experienced practitioner to identify the precise outline of the structures of interest. This complicates, delays, and increases the cost of network training and deployment. To address this problem, we propose a multi-path decoder U-Net architecture that is trained on bounding box segmentation maps; not requiring pixel-wise annotations. We show that the network can be trained on small training sets, which is the case in medical imaging datasets; reducing the cost and time needed for deployment and use in clinical settings. The multi-path decoder design allows for better training of deeper layers and earlier attention to the target anatomical structures of interest. This architecture offers up to a 7% relative improvement compared to the U-Net architecture in localization and detection performance, with an increase of only 0.75% in the number of parameters. Its performance is on par with, or slightly better than, the more computationally expensive U-Net++, which has 20% more parameters; making the proposed architecture a more computationally efficient alternative for real-time object detection and localization in ultrasound scans.

Nucleus segmentation and classification using residual SE-UNet and feature concatenation approach incervical cytopathology cell images

Introduction: Pap smear is considered to be the primary examination for the diagnosis of cervical cancer. But the analysis of pap smear slides is a time-consuming task and tedious as it requires manual intervention. The diagnostic efficiency depends on the medical expertise of the pathologist, and human error often hinders the diagnosis. Automated segmentation and classification of cervical nuclei will help diagnose cervical cancer in earlier stages. Materials and Methods: The proposed methodology includes three models: a Residual-Squeeze-and-Excitation-module based segmentation model, a fusion-based feature extraction model, and a Multi-layer Perceptron classification model. In the fusion-based feature extraction model, three sets of deep features are extracted from these segmented nuclei using the pre-trained and fine-tuned VGG19, VGG-F, and CaffeNet models, and two hand-crafted descriptors, Bag-of-Features and Linear-Binary-Patterns, are extracted for each image. For this work, Herlev, SIPaKMeD, and ISBI2014 datasets are used for evaluation. The Herlev datasetis used for evaluating both segmentation and classification models. Whereas the SIPaKMeD and ISBI2014 are used for evaluating the classification model, and the segmentation model respectively. Results: The segmentation network enhanced the precision and ZSI by 2.04%, and 2.00% on the Herlev dataset, and the precision and recall by 0.68%, and 2.59% on the ISBI2014 dataset. The classification approach enhanced the accuracy, recall, and specificity by 0.59%, 0.47%, and 1.15% on the Herlev dataset, and by 0.02%, 0.15%, and 0.22% on the SIPaKMed dataset. Conclusion: The experiments demonstrate that the proposed work achieves promising performance on segmentation and classification in cervical cytopathology cell images..

Non-Expert Markings of Active Chronic Graft-Versus-Host Disease Photographs: Optimal Metrics of Training Effects

Lack of reliable measures of cutaneous chronic graft-versus-host disease (cGVHD) remains a significant challenge. Non-expert assistance in marking photographs of active disease could aid the development of automated segmentation algorithms, but validated metrics to evaluate training effects are lacking. We studied absolute and relative error of marked body surface area (BSA), redness, and the Dice index as potential metrics of non-expert improvement. Three non-experts underwent an extensive training program led by a board-certified dermatologist to mark cGVHD in photographs. At the end of the 4-month training, the dermatologist confirmed that each trainee had learned to accurately mark cGVHD. The trainees' inter- and intra-rater intraclass correlation coefficient estimates were "substantial" to "almost perfect" for both BSA and total redness. For fifteen 3D photos of patients with cGVHD, the trainees' median absolute (relative) BSA error compared to expert marking dropped from 20 cm2 (29%) pre-training to 14 cm2 (24%) post-training. Total redness error decreased from 122 a*·cm2 (26%) to 95 a*·cm2 (21%). By contrast, median Dice index did not reflect improvement (0.76 to 0.75). Both absolute and relative BSA and redness errors similarly and stably reflected improvements from this training program, which the Dice index failed to capture.

The top 100 most cited articles on artificial intelligence in radiology: a bibliometric analysis

AIM

To identify the most influential publications relating to artificial intelligence (AI) in radiology in order to identify current trends in the literature and to highlight areas requiring further research.

MATERIALS AND METHODS

A retrospective bibliometric analysis was performed of the top 100 most cited articles on this topic. Data pertaining to year of publication, publishing journal, journal impact factor, authorship, article title, institution, country, type of article, article subject, and keywords were collected.

RESULTS

The number of citations per article for the top 100 list ranged from 254 to 3,576 (median 353). The number of citations per year, per article ranged from 10.4 to 894 (median 65.6). The majority of articles (n=62) were published within the last 10 years. The USA was the most common country of origin (n=44). The journal with the greatest number of articles was IEEE Transactions On Medical Imaging (n=38). University Medical Center Utrecht contributed the greatest number of articles (n=6). There were 92 original research articles, 52 of which were clinical studies. The most common clinical subjects were neuroimaging (n=25) and oncology (n=16). The most common keyword used was "deep learning" (n=34).

CONCLUSION

This study provides an in-depth analysis of the top 100 most-cited papers on the use of AI in radiology. It also provides researchers with detailed insight into the current influential papers in this field, the characteristics of those studies, as well as potential future trends in this fast-developing area of radiology.

Pipeline for Automated Processing of Clinical Cone-Beam Computed Tomography for Patient-Specific Temporal Bone Simulation: Validation and Clinical Feasibility

OBJECTIVE

Patient-specific simulation allows the surgeon to plan and rehearse the surgical approach ahead of time. Preoperative clinical imaging for this purpose requires time-consuming manual processing and segmentation of landmarks such as the facial nerve. We aimed to evaluate an automated pipeline with minimal manual interaction for processing clinical cone-beam computed tomography (CBCT) temporal bone imaging for patient-specific virtual reality (VR) simulation.

STUDY DESIGN

Prospective image processing of retrospective imaging series.

SETTING

Academic hospital.

METHODS

Eleven CBCTs were selected based on quality and used for validation of the processing pipeline. A larger naturalistic sample of 36 CBCTs were obtained to explore parameters for successful processing and feasibility for patient-specific VR simulation.Visual inspection and quantitative metrics were used to validate the accuracy of automated segmentation compared with manual segmentation. Range of acceptable rotational offsets and translation point selection variability were determined. Finally, feasibility in relation to image acquisition quality, processing time, and suitability for VR simulation was evaluated.

RESULTS

The performance of automated segmentation was acceptable compared with manual segmentation as reflected in the quantitative metrics. Total time for processing for new data sets was on average 8.3 minutes per data set; of this, it was less than 30 seconds for manual steps. Two of the 36 data sets failed because of extreme rotational offset, but overall the registration routine was robust to rotation and manual selection of a translational reference point. Another seven data sets had successful automated segmentation but insufficient suitability for VR simulation.

CONCLUSION

Automated processing of CBCT imaging has potential for preoperative VR simulation but requires further refinement.

Real-time image-guided treatment of mobile tumors in proton therapy by a library of treatment plans: a simulation study

PURPOSE

To improve target coverage and reduce the dose in the surrounding organs-at-risks (OARs), we developed an image-guided treatment method based on a precomputed library of treatment plans controlled and delivered in real-time.

METHODS

A library of treatment plans is constructed by optimizing a plan for each breathing phase of a four dimensional computed tomography (4DCT). Treatments are delivered by simulation on a continuous sequence of synthetic computed tomographies (CTs) generated from real magnetic resonance imaging (MRI) sequences. During treatment, the plans for which the tumor are at a close distance to the current tumor position are selected to deliver their spots. The study is conducted on five liver cases.

RESULTS

We tested our approach under imperfect knowledge of the tumor positions with a 2 mm distance error. On average, compared to a 4D robustly optimized treatment plan, our approach led to a dose homogeneity increase of 5% (defined as 1 - D 5 - D 95 prescription $1-\frac{D_5-D_{95}}{\text{prescription}}$ ) in the target and a mean liver dose decrease of 23%. The treatment time was roughly increased by a factor of 2 but remained below 4 min on average.

CONCLUSIONS

Our image-guided treatment framework outperforms state-of-the-art 4D-robust plans for all patients in this study on both target coverage and OARs sparing, with an acceptable increase in treatment time under the current accuracy of the tumor tracking technology.

Machine learning segmentation of core and penumbra from acute stroke CT perfusion data

Introduction

Computed tomography perfusion (CTP) imaging is widely used in cases of suspected acute ischemic stroke to positively identify ischemia and assess suitability for treatment through identification of reversible and irreversible tissue injury. Traditionally, this has been done via setting single perfusion thresholds on two or four CTP parameter maps. We present an alternative model for the estimation of tissue fate using multiple perfusion measures simultaneously.

Methods

We used machine learning (ML) models based on four different algorithms, combining four CTP measures (cerebral blood flow, cerebral blood volume, mean transit time and delay time) plus 3D-neighborhood (patch) analysis to predict the acute ischemic core and perfusion lesion volumes. The model was developed using 86 patient images, and then tested further on 22 images.

Results

XGBoost was the highest-performing algorithm. With standard threshold-based core and penumbra measures as the reference, the model demonstrated moderate agreement in segmenting core and penumbra on test images. Dice similarity coefficients for core and penumbra were 0.38 ± 0.26 and 0.50 ± 0.21, respectively, demonstrating moderate agreement. Skull-related image artefacts contributed to lower accuracy.

Discussion

Further development may enable us to move beyond the current overly simplistic core and penumbra definitions using single thresholds where a single error or artefact may lead to substantial error.

Prognosis prediction of uterine cervical cancer using changes in the histogram and texture features of apparent diffusion coefficient during definitive chemoradiotherapy

OBJECTIVES

We investigated prospectively whether, in cervical cancer (CC) treated with concurrent chemoradiotherapy (CCRT), the Apparent diffusion coefficient (ADC) histogram and texture parameters and their change rates during treatment could predict prognosis.

METHODS

Fifty-seven CC patients treated with CCRT at our institution were included. They underwent MRI scans up to four times during the treatment course (1st, before treatment [n = 41], 2nd, at the start of image-guided brachytherapy (IGBT) [n = 41], 3rd, in the middle of IGBT [n = 27], 4th, after treatment [n = 53]). The entire tumor was manually set as the volume of interest (VOI) manually in the axial images of the ADC map by two radiologists. A total of 107 image features (morphology features 14, histogram features 18, texture features 75) were extracted from the VOI. The recurrence prediction values of the features and their change rates were evaluated by Receiver operating characteristics (ROC) analysis. The presence or absence of local and distant recurrence within two years was set as an outcome. The intraclass correlation coefficient (ICC) was also calculated.

RESULTS

The change rates in kurtosis between the 1st and 3rd, and 1st and 2nd MRIs, and the change rate in grey level co-occurrence matrix_cluster shade between the 2nd and 3rd MRIs showed particularly high predictive powers (area under the ROC curve = 0.785, 0.759, and 0.750, respectively), which exceeded the predictive abilities of the parameters obtained from pre- or post-treatment MRI only. The change rate in kurtosis between the 1st and 2nd MRIs had good reliability (ICC = 0.765).

CONCLUSIONS

The change rate in ADC kurtosis between the 1st and 2nd MRIs was the most reliable parameter, enabling us to predict prognosis early in the treatment course.