The Liver Tumor Segmentation Benchmark (LiTS).

In this work, we report the set-up and results of the Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI) 2017 and the International Conferences on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2017 and 2018. The image dataset is diverse and contains primary and secondary tumors with varied sizes and appearances with various lesion-to-background levels (hyper-/hypo-dense), created in collaboration with seven hospitals and research institutions. Seventy-five submitted liver and liver tumor segmentation algorithms were trained on a set of 131 computed tomography (CT) volumes and were tested on 70 unseen test images acquired from different patients. We found that not a single algorithm performed best for both liver and liver tumors in the three events. The best liver segmentation algorithm achieved a Dice score of 0.963, whereas, for tumor segmentation, the best algorithms achieved Dices scores of 0.674 (ISBI 2017), 0.702 (MICCAI 2017), and 0.739 (MICCAI 2018). Retrospectively, we performed additional analysis on liver tumor detection and revealed that not all top-performing segmentation algorithms worked well for tumor detection. The best liver tumor detection method achieved a lesion-wise recall of 0.458 (ISBI 2017), 0.515 (MICCAI 2017), and 0.554 (MICCAI 2018), indicating the need for further research. LiTS remains an active benchmark and resource for research, e.g., contributing the liver-related segmentation tasks in http://medicaldecathlon.com/. In addition, both data and online evaluation are accessible via https://competitions.codalab.org/competitions/17094.

Automated Detection of Brain Metastases on T1-Weighted MRI Using a Convolutional Neural Network: Impact of Volume Aware Loss and Sampling Strategy.

BACKGROUND: Detection of brain metastases (BM) and segmentation for treatment planning could be optimized with machine learning methods. Convolutional neural networks (CNNs) are promising, but their trade-offs between sensitivity and precision frequently lead to missing small lesions. HYPOTHESIS: Combining volume aware (VA) loss function and sampling strategy could improve BM detection sensitivity. STUDY TYPE: Retrospective. POPULATION: A total of 530 radiation oncology patients (55% women) were split into a training/validation set (433 patients/1460 BM) and an independent test set (97 patients/296 BM). FIELD STRENGTH/SEQUENCE: 1.5 T and 3 T, contrast-enhanced three-dimensional (3D) T1-weighted fast gradient echo sequences. ASSESSMENT: Ground truth masks were based on radiotherapy treatment planning contours reviewed by experts. A U-Net inspired model was trained. Three loss functions (Dice, Dice + boundary, and VA) and two sampling methods (label and VA) were compared. Results were reported with Dice scores, volumetric error, lesion detection sensitivity, and precision. A detected voxel within the ground truth constituted a true positive. STATISTICAL TESTS: McNemar's exact test to compare detected lesions between models. Pearson's correlation coefficient and Bland-Altman analysis to compare volume agreement between predicted and ground truth volumes. Statistical significance was set at P ≤ 0.05. RESULTS: Combining VA loss and VA sampling performed best with an overall sensitivity of 91% and precision of 81%. For BM in the 2.5-6 mm estimated sphere diameter range, VA loss reduced false negatives by 58% and VA sampling reduced it further by 30%. In the same range, the boundary loss achieved the highest precision at 81%, but a low sensitivity (24%) and a 31% Dice loss. DATA CONCLUSION: Considering BM size in the loss and sampling function of CNN may increase the detection sensitivity regarding small BM. Our pipeline relying on a single contrast-enhanced T1-weighted MRI sequence could reach a detection sensitivity of 91%, with an average of only 0.66 false positives per scan. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 2.

Deep Learning: An Update for Radiologists.

Deep learning is a class of machine learning methods that has been successful in computer vision. Unlike traditional machine learning methods that require hand-engineered feature extraction from input images, deep learning methods learn the image features by which to classify data. Convolutional neural networks (CNNs), the core of deep learning methods for imaging, are multilayered artificial neural networks with weighted connections between neurons that are iteratively adjusted through repeated exposure to training data. These networks have numerous applications in radiology, particularly in image classification, object detection, semantic segmentation, and instance segmentation. The authors provide an update on a recent primer on deep learning for radiologists, and they review terminology, data requirements, and recent trends in the design of CNNs; illustrate building blocks and architectures adapted to computer vision tasks, including generative architectures; and discuss training and validation, performance metrics, visualization, and future directions. Familiarity with the key concepts described will help radiologists understand advances of deep learning in medical imaging and facilitate clinical adoption of these techniques. Online supplemental material is available for this article. ©RSNA, 2021.

Learning normalized inputs for iterative estimation in medical image segmentation.

In this paper, we introduce a simple, yet powerful pipeline for medical image segmentation that combines Fully Convolutional Networks (FCNs) with Fully Convolutional Residual Networks (FC-ResNets). We propose and examine a design that takes particular advantage of recent advances in the understanding of both Convolutional Neural Networks as well as ResNets. Our approach focuses upon the importance of a trainable pre-processing when using FC-ResNets and we show that a low-capacity FCN model can serve as a pre-processor to normalize medical input data. In our image segmentation pipeline, we use FCNs to obtain normalized images, which are then iteratively refined by means of a FC-ResNet to generate a segmentation prediction. As in other fully convolutional approaches, our pipeline can be used off-the-shelf on different image modalities. We show that using this pipeline, we exhibit state-of-the-art performance on the challenging Electron Microscopy benchmark, when compared to other 2D methods. We improve segmentation results on CT images of liver lesions, when contrasting with standard FCN methods. Moreover, when applying our 2D pipeline on a challenging 3D MRI prostate segmentation challenge we reach results that are competitive even when compared to 3D methods. The obtained results illustrate the strong potential and versatility of the pipeline by achieving accurate segmentations on a variety of image modalities and different anatomical regions.

Deep Learning: A Primer for Radiologists.

Deep learning is a class of machine learning methods that are gaining success and attracting interest in many domains, including computer vision, speech recognition, natural language processing, and playing games. Deep learning methods produce a mapping from raw inputs to desired outputs (eg, image classes). Unlike traditional machine learning methods, which require hand-engineered feature extraction from inputs, deep learning methods learn these features directly from data. With the advent of large datasets and increased computing power, these methods can produce models with exceptional performance. These models are multilayer artificial neural networks, loosely inspired by biologic neural systems. Weighted connections between nodes (neurons) in the network are iteratively adjusted based on example pairs of inputs and target outputs by back-propagating a corrective error signal through the network. For computer vision tasks, convolutional neural networks (CNNs) have proven to be effective. Recently, several clinical applications of CNNs have been proposed and studied in radiology for classification, detection, and segmentation tasks. This article reviews the key concepts of deep learning for clinical radiologists, discusses technical requirements, describes emerging applications in clinical radiology, and outlines limitations and future directions in this field. Radiologists should become familiar with the principles and potential applications of deep learning in medical imaging. ©RSNA, 2017.

Liver segmentation: indications, techniques and future directions.

OBJECTIVES: Liver volumetry has emerged as an important tool in clinical practice. Liver volume is assessed primarily via organ segmentation of computed tomography (CT) and magnetic resonance imaging (MRI) images. The goal of this paper is to provide an accessible overview of liver segmentation targeted at radiologists and other healthcare professionals. METHODS: Using images from CT and MRI, this paper reviews the indications for liver segmentation, technical approaches used in segmentation software and the developing roles of liver segmentation in clinical practice. RESULTS: Liver segmentation for volumetric assessment is indicated prior to major hepatectomy, portal vein embolisation, associating liver partition and portal vein ligation for staged hepatectomy (ALPPS) and transplant. Segmentation software can be categorised according to amount of user input involved: manual, semi-automated and fully automated. Manual segmentation is considered the "gold standard" in clinical practice and research, but is tedious and time-consuming. Increasingly automated segmentation approaches are more robust, but may suffer from certain segmentation pitfalls. Emerging applications of segmentation include surgical planning and integration with MRI-based biomarkers. CONCLUSIONS: Liver segmentation has multiple clinical applications and is expanding in scope. Clinicians can employ semi-automated or fully automated segmentation options to more efficiently integrate volumetry into clinical practice. TEACHING POINTS: • Liver volume is assessed via organ segmentation on CT and MRI examinations. • Liver segmentation is used for volume assessment prior to major hepatic procedures. • Segmentation approaches may be categorised according to the amount of user input involved. • Emerging applications include surgical planning and integration with MRI-based biomarkers.

Liver Segmentation on CT and MR Using Laplacian Mesh Optimization.

OBJECTIVE: The purpose of this paper is to describe a semiautomated segmentation method for the liver and evaluate its performance on CT-scan and MR images. METHODS: First, an approximate 3-D model of the liver is initialized from a few user-generated contours to globally outline the liver shape. The model is then automatically deformed by a Laplacian mesh optimization scheme until it precisely delineates the patient's liver. A correction tool was implemented to allow the user to improve the segmentation until satisfaction. RESULTS: The proposed method was tested against 30 CT-scans from the SLIVER07 challenge repository and 20 MR studies from the Montreal University Hospital Center, covering a wide spectrum of liver morphologies and pathologies. The average volumetric overlap error was 5.1% for CT and 7.6% for MRI and the average segmentation time was 6 min. CONCLUSION: The obtained results show that the proposed method is efficient, reliable, and could effectively be used routinely in the clinical setting. SIGNIFICANCE: The proposed approach can alleviate the cumbersome and tedious process of slice-wise segmentation required for precise hepatic volumetry, virtual surgery, and treatment planning.

Comparison of MRI- and CT-based semiautomated liver segmentation: a validation study.

PURPOSE: To compare the repeatability, agreement, and efficiency of MRI- and CT-based semiautomated liver segmentation for the assessment of total and subsegmental liver volume. METHODS: This retrospective study was conducted in 31 subjects who underwent contemporaneous liver MRI and CT. Total and subsegmental liver volumes were segmented from contrast-enhanced 3D gradient-recalled echo MRI sequences and CT images. Semiautomated segmentation was based on variational interpolation and Laplacian mesh optimization. All segmentations were repeated after 2 weeks. Manual segmentation of CT images using an active contour tool was used as the reference standard. Repeatability and agreement of the methods were evaluated with intra-class correlation coefficients (ICC) and Bland-Altman analysis. Total interaction time was recorded. RESULTS: Intra-reader ICC were ≥0.987 for MRI and ≥0.995 for CT. Intra-reader repeatability was 30 ± 217 ml (bias ± 1.96 SD) (95% limits of agreement: -187 to 247 ml) for MRI and -10 ± 143 ml (-153 to 133 ml) for CT. Inter-method ICC between semiautomated and manual volumetry were ≥0.995 for MRI and ≥0.986 for CT. Inter-method segmental ICC varied between 0.584 and 0.865 for MRI and between 0.596 and 0.890 for CT. Inter-method agreement was -14 ± 136 ml (-150 to 122 ml) for MRI and 50 ± 226 ml (-176 to 276 ml) for CT. Inter-method segmental agreement ranged from 10 ± 47 ml (-37 to 57 ml) to 2 ± 214 ml (-212 to 216 ml) for MRI and 9 ± 45 ml (-36 to 54 ml) to -46 ± 183 ml (-229 to 137 ml) for CT. Interaction time (mean ± SD) was significantly shorter for MRI-based semiautomated segmentation (7.2 ± 0.1 min, p < 0.001) and for CT-based semiautomated segmentation (6.5 ± 0.2 min, p < 0.001) than for CT-based manual segmentation (14.5 ± 0.4 min). CONCLUSION: MRI-based semiautomated segmentation provides similar repeatability and agreement to CT-based segmentation for total liver volume.

MRI-determined liver proton density fat fraction, with MRS validation: Comparison of regions of interest sampling methods in patients with type 2 diabetes.

PURPOSE: To assess the agreement between published magnetic resonance imaging (MRI)-based regions of interest (ROI) sampling methods using liver mean proton density fat fraction (PDFF) as the reference standard. MATERIALS AND METHODS: This retrospective, internal review board-approved study was conducted in 35 patients with type 2 diabetes. Liver PDFF was measured by magnetic resonance spectroscopy (MRS) using a stimulated-echo acquisition mode sequence and MRI using a multiecho spoiled gradient-recalled echo sequence at 3.0T. ROI sampling methods reported in the literature were reproduced and liver mean PDFF obtained by whole-liver segmentation was used as the reference standard. Intraclass correlation coefficients (ICCs), Bland-Altman analysis, repeated-measures analysis of variance (ANOVA), and paired t-tests were performed. RESULTS: ICC between MRS and MRI-PDFF was 0.916. Bland-Altman analysis showed excellent intermethod agreement with a bias of -1.5 ± 2.8%. The repeated-measures ANOVA found no systematic variation of PDFF among the nine liver segments. The correlation between liver mean PDFF and ROI sampling methods was very good to excellent (0.873 to 0.975). Paired t-tests revealed significant differences (P < 0.05) with ROI sampling methods that exclusively or predominantly sampled the right lobe. Significant correlations with mean PDFF were found with sampling methods that included higher number of segments, total area equal or larger than 5 cm(2) , or sampled both lobes (P = 0.001, 0.023, and 0.002, respectively). CONCLUSION: MRI-PDFF quantification methods should sample each liver segment in both lobes and include a total surface area equal or larger than 5 cm(2) to provide a close estimate of the liver mean PDFF.

[(18)F]-fluorodeoxyglucose positron emission tomography of the cat brain: A feasibility study to investigate osteoarthritis-associated pain.

The objective of this pilot study was to investigate central nervous system (CNS) changes related to osteoarthritis (OA)-associated chronic pain in cats using [(18)F]-fluorodeoxyglucose ((18)FDG) positron emission tomography (PET) imaging. The brains of five normal, healthy (non-OA) cats and seven cats with pain associated with naturally occurring OA were imaged using (18)FDG-PET during a standardized mild anesthesia protocol. The PET images were co-registered over a magnetic resonance image of a cat brain segmented into several regions of interest. Brain metabolism was assessed in these regions using standardized uptake values. The brain metabolism in the secondary somatosensory cortex, thalamus and periaqueductal gray matter was increased significantly (P ≤ 0.005) in OA cats compared with non-OA cats. This study indicates that (18)FDG-PET brain imaging in cats is feasible to investigate CNS changes related to chronic pain. The results also suggest that OA is associated with sustained nociceptive inputs and increased activity of the descending modulatory pathways.

Validation of a semiautomated liver segmentation method using CT for accurate volumetry.

RATIONALE AND OBJECTIVES: To compare the repeatability and agreement of a semiautomated liver segmentation method with manual segmentation for assessment of total liver volume on CT (computed tomography). MATERIALS AND METHODS: This retrospective, institutional review board-approved study was conducted in 41 subjects who underwent liver CT for preoperative planning. The major pathologies encountered were colorectal cancer metastases, benign liver lesions and hepatocellular carcinoma. This semiautomated segmentation method is based on variational interpolation and 3D minimal path-surface segmentation. Total and subsegmental liver volumes were segmented from contrast-enhanced CT images in venous phase. Two image analysts independently performed semiautomated segmentations and two other image analysts performed manual segmentations. Repeatability and agreement of both methods were evaluated with intraclass correlation coefficients (ICC) and Bland-Altman analysis. Interaction time was recorded for both methods. RESULTS: Bland-Altman analysis revealed an intrareader agreement of -1 ± 27 mL (mean ± 1.96 standard deviation) with ICC of 0.999 (P < .001) for manual segmentation and 12 ± 97 mL with ICC of 0.991 (P < .001) for semiautomated segmentation. Bland-Altman analysis revealed an interreader agreement of -4 ± 22 mL with ICC of 0.999 (P < .001) for manual segmentation and 5 ± 98 mL with ICC of 0.991 (P < .001) for semiautomated segmentation. Intermethod agreement was found to be 3 ± 120 mL with ICC of 0.988 (P < .001). Mean interaction time was 34.3 ± 16.7 minutes for the manual method and 8.0 ± 1.2 minutes for the semiautomated method (P < .001). CONCLUSIONS: A semiautomated segmentation method can substantially shorten interaction time while preserving a high repeatability and agreement with manual segmentation.

Effects of Insulin Glargine and Liraglutide Therapy on Liver Fat as Measured by Magnetic Resonance in Patients With Type 2 Diabetes: A Randomized Trial.

OBJECTIVE: This study determined the effects of insulin versus liraglutide therapy on liver fat in patients with type 2 diabetes inadequately controlled with oral agents therapy, including metformin. RESEARCH DESIGN AND METHODS: Thirty-five patients with type 2 diabetes inadequately controlled on metformin monotherapy or in combination with other oral antidiabetic medications were randomized to receive insulin glargine or liraglutide therapy for 12 weeks. The liver proton density fat fraction (PDFF) was measured by MRS. The mean liver PDFF, the total liver volume, and the total liver fat index were measured by MRI. The Student t test, the Fisher exact test, and repeated-measures ANOVA were used for statistical analysis. RESULTS: Insulin treatment was associated with a significant improvement in glycated hemoglobin (7.9% to 7.2% [62.5 to 55.2 mmol/mol], P = 0.005), a trend toward a decrease in MRS-PDFF (12.6% to 9.9%, P = 0.06), and a significant decrease in liver mean MRI-PDFF (13.8% to 10.6%, P = 0.005), liver volume (2,010.6 to 1,858.7 mL, P = 0.01), and the total liver fat index (304.4 vs. 209.3 % ⋅ mL, P = 0.01). Liraglutide treatment was also associated with a significant improvement in glycated hemoglobin (7.6% to 6.7% [59.8 to 50.2 mmol/mol], P < 0.001) but did not change MRS-PDFF (P = 0.80), liver mean MRI-PDFF (P = 0.15), liver volume (P = 0.30), or the total liver fat index (P = 0.39). CONCLUSIONS: The administration of insulin glargine therapy reduced the liver fat burden in patients with type 2 diabetes. However, the improvements in the liver fat fraction and glycemia control were not significantly different from those in the liraglutide group.