Deep Learning-Based Assessment of Functional Liver Capacity Using Gadoxetic Acid-Enhanced Hepatobiliary Phase MRI.

OBJECTIVE: We aimed to develop and test a deep learning algorithm (DLA) for fully automated measurement of the volume and signal intensity (SI) of the liver and spleen using gadoxetic acid-enhanced hepatobiliary phase (HBP)-magnetic resonance imaging (MRI) and to evaluate the clinical utility of DLA-assisted assessment of functional liver capacity. MATERIALS AND METHODS: The DLA was developed using HBP-MRI data from 1014 patients. Using an independent test dataset (110 internal and 90 external MRI data), the segmentation performance of the DLA was measured using the Dice similarity score (DSS), and the agreement between the DLA and the ground truth for the volume and SI measurements was assessed with a Bland-Altman 95% limit of agreement (LOA). In 276 separate patients (male:female, 191:85; mean age ± standard deviation, 40 ± 15 years) who underwent hepatic resection, we evaluated the correlations between various DLA-based MRI indices, including liver volume normalized by body surface area (LVBSA), liver-to-spleen SI ratio (LSSR), MRI parameter-adjusted LSSR (aLSSR), LSSR × LVBSA, and aLSSR × LVBSA, and the indocyanine green retention rate at 15 minutes (ICG-R15), and determined the diagnostic performance of the DLA-based MRI indices to detect ICG-R15 ≥ 20%. RESULTS: In the test dataset, the mean DSS was 0.977 for liver segmentation and 0.946 for spleen segmentation. The Bland-Altman 95% LOAs were 0.08% ± 3.70% for the liver volume, 0.20% ± 7.89% for the spleen volume, -0.02% ± 1.28% for the liver SI, and -0.01% ± 1.70% for the spleen SI. Among DLA-based MRI indices, aLSSR × LVBSA showed the strongest correlation with ICG-R15 (r = -0.54, p < 0.001), with area under receiver operating characteristic curve of 0.932 (95% confidence interval, 0.895-0.959) to diagnose ICG-R15 ≥ 20%. CONCLUSION: Our DLA can accurately measure the volume and SI of the liver and spleen and may be useful for assessing functional liver capacity using gadoxetic acid-enhanced HBP-MRI.

Development of a fully automatic deep learning system for L3 selection and body composition assessment on computed tomography.

As sarcopenia research has been gaining emphasis, the need for quantification of abdominal muscle on computed tomography (CT) is increasing. Thus, a fully automated system to select L3 slice and segment muscle in an end-to-end manner is demanded. We aimed to develop a deep learning model (DLM) to select the L3 slice with consideration of anatomic variations and to segment cross-sectional areas (CSAs) of abdominal muscle and fat. Our DLM, named L3SEG-net, was composed of a YOLOv3-based algorithm for selecting the L3 slice and a fully convolutional network (FCN)-based algorithm for segmentation. The YOLOv3-based algorithm was developed via supervised learning using a training dataset (n = 922), and the FCN-based algorithm was transferred from prior work. Our L3SEG-net was validated with internal (n = 496) and external validation (n = 586) datasets. Ground truth L3 level CT slice and anatomic variation were identified by a board-certified radiologist. L3 slice selection accuracy was evaluated by the distance difference between ground truths and DLM-derived results. Technical success for L3 slice selection was defined when the distance difference was < 10 mm. Overall segmentation accuracy was evaluated by CSA error and DSC value. The influence of anatomic variations on DLM performance was evaluated. In the internal and external validation datasets, the accuracy of automatic L3 slice selection was high, with mean distance differences of 3.7 ± 8.4 mm and 4.1 ± 8.3 mm, respectively, and with technical success rates of 93.1% and 92.3%, respectively. However, in the subgroup analysis of anatomic variations, the L3 slice selection accuracy decreased, with distance differences of 12.4 ± 15.4 mm and 12.1 ± 14.6 mm, respectively, and with technical success rates of 67.2% and 67.9%, respectively. The overall segmentation accuracy of abdominal muscle areas was excellent regardless of anatomic variation, with CSA errors of 1.38-3.10 cm2. A fully automatic system was developed for the selection of an exact axial CT slice at the L3 vertebral level and the segmentation of abdominal muscle areas.

CDISC-compliant clinical trial imaging management system with automatic verification and data Transformation: Focusing on tumor response assessment data in clinical trials.

OBJECTIVE: Major issues in imaging data management of tumor response assessment in clinical trials include high human errors in data input and unstandardized data structures, warranting a new breakthrough IT solution. Thus, we aim to develop a Clinical Data Interchange Standards Consortium (CDISC)-compliant clinical trial imaging management system (CTIMS) with automatic verification and transformation modules for implementing the CDISC Study Data Tabulation Model (SDTM) in the tumor response assessment dataset of clinical trials. MATERIALS AND METHODS: In accordance with various CDISC standards guides and Response Evaluation Criteria in Solid Tumors (RECIST) guidelines, the overall system architecture of CDISC-compliant CTIMS was designed. Modules for standard-compliant electronic case report form (eCRF) to verify data conformance and transform into SDTM data format were developed by experts in diverse fields such as medical informatics, medical, and clinical trial. External validation of the CDISC-compliant CTIMS was performed by comparing it with our previous CTIMS based on real-world data and CDISC validation rules by Pinnacle 21 Community Software. RESULTS: The architecture of CDISC-compliant CTIMS included the standard-compliant eCRF module of RECIST, the automatic verification module of the input data, and the SDTM transformation module from the eCRF input data to the SDTM datasets based on CDISC Define-XML. This new system was incorporated into our previous CTIMS. External validation demonstrated that all 176 human input errors occurred in the previous CTIMS filtered by a new system yielding zero error and CDISC-compliant dataset. The verified eCRF input data were automatically transformed into the SDTM dataset, which satisfied the CDISC validation rules by Pinnacle 21 Community Software. CONCLUSIONS: To assure data consistency and high quality of the tumor response assessment data, our new CTIMS can minimize human input error by using standard-compliant eCRF with an automatic verification module and automatically transform the datasets into CDISC SDTM format.

MRI Predictors of Malignant Transformation in Patients with Inverted Papilloma: A Decision Tree Analysis Using Conventional Imaging Features and Histogram Analysis of Apparent Diffusion Coefficients.

OBJECTIVE: Preoperative differentiation between inverted papilloma (IP) and its malignant transformation to squamous cell carcinoma (IP-SCC) is critical for patient management. We aimed to determine the diagnostic accuracy of conventional imaging features and histogram parameters obtained from whole tumor apparent diffusion coefficient (ADC) values to predict IP-SCC in patients with IP, using decision tree analysis. MATERIALS AND METHODS: In this retrospective study, we analyzed data generated from the records of 180 consecutive patients with histopathologically diagnosed IP or IP-SCC who underwent head and neck magnetic resonance imaging, including diffusion-weighted imaging and 62 patients were included in the study. To obtain whole tumor ADC values, the region of interest was placed to cover the entire volume of the tumor. Classification and regression tree analyses were performed to determine the most significant predictors of IP-SCC among multiple covariates. The final tree was selected by cross-validation pruning based on minimal error. RESULTS: Of 62 patients with IP, 21 (34%) had IP-SCC. The decision tree analysis revealed that the loss of convoluted cerebriform pattern and the 20th percentile cutoff of ADC were the most significant predictors of IP-SCC. With these decision trees, the sensitivity, specificity, accuracy, and C-statistics were 86% (18 out of 21; 95% confidence interval [CI], 65-95%), 100% (41 out of 41; 95% CI, 91-100%), 95% (59 out of 61; 95% CI, 87-98%), and 0.966 (95% CI, 0.912-1.000), respectively. CONCLUSION: Decision tree analysis using conventional imaging features and histogram analysis of whole volume ADC could predict IP-SCC in patients with IP with high diagnostic accuracy.

An index based on deep learning-measured spleen volume on CT for the assessment of high-risk varix in B-viral compensated cirrhosis.

OBJECTIVES: Deep learning enables an automated liver and spleen volume measurements on CT. The purpose of this study was to develop an index combining liver and spleen volumes and clinical factors for detecting high-risk varices in B-viral compensated cirrhosis. METHODS: This retrospective study included 419 patients with B-viral compensated cirrhosis who underwent endoscopy and CT from 2007 to 2008 (derivation cohort, n = 239) and from 2009 to 2010 (validation cohort, n = 180). The liver and spleen volumes were measured on CT images using a deep learning algorithm. Multivariable logistic regression analysis of the derivation cohort developed an index to detect endoscopically confirmed high-risk varix. The cumulative 5-year risk of varix bleeding was evaluated with patients stratified by their index values. RESULTS: The index of spleen volume-to-platelet ratio was devised from the derivation cohort. In the validation cohort, the cutoff index value for balanced sensitivity and specificity (> 3.78) resulted in the sensitivity of 69.4% and the specificity of 78.5% for detecting high-risk varix, and the cutoff index value for high sensitivity (> 1.63) detected all high-risk varices. The index stratified all patients into the low (index value ≤ 1.63; n = 118), intermediate (n = 162), and high (index value > 3.78; n = 139) risk groups with cumulative 5-year incidences of varix bleeding of 0%, 1.0%, and 12.0%, respectively (p < .001). CONCLUSION: The spleen volume-to-platelet ratio obtained using deep learning-based CT analysis is useful to detect high-risk varices and to assess the risk of varix bleeding. KEY POINTS: • The criterion of spleen volume to platelet > 1.63 detected all high-risk varices in the validation cohort, while the absence of visible varix did not exclude all high-risk varices. • Visual varix grade ≥ 2 detected high-risk varix with a high specificity (96.5-100%). • Combining spleen volume-to-platelet ratio ≤ 1.63 and visual varix grade of 0 identified low-risk patients who had no high-risk varix and varix bleeding on 5-year follow-up.

Assessment of Myosteatosis on Computed Tomography by Automatic Generation of a Muscle Quality Map Using a Web-Based Toolkit: Feasibility Study.

BACKGROUND: Muscle quality is associated with fatty degeneration or infiltration of the muscle, which may be associated with decreased muscle function and increased disability. OBJECTIVE: The aim of this study is to evaluate the feasibility of automated quantitative measurements of the skeletal muscle on computed tomography (CT) images to assess normal-attenuation muscle and myosteatosis. METHODS: We developed a web-based toolkit to generate a muscle quality map by categorizing muscle components. First, automatic segmentation of the total abdominal muscle area (TAMA), visceral fat area, and subcutaneous fat area was performed using a predeveloped deep learning model on a single axial CT image at the L3 vertebral level. Second, the Hounsfield unit of each pixel in the TAMA was measured and categorized into 3 components: normal-attenuation muscle area (NAMA), low-attenuation muscle area (LAMA), and inter/intramuscular adipose tissue (IMAT) area. The myosteatosis area was derived by adding the LAMA and IMAT area. We tested the feasibility of the toolkit using randomly selected healthy participants, comprising 6 different age groups (20 to 79 years). With stratification by sex, these indices were compared between age groups using 1-way analysis of variance (ANOVA). Correlations between the myosteatosis area or muscle densities and fat areas were analyzed using Pearson correlation coefficient r. RESULTS: A total of 240 healthy participants (135 men and 105 women) with 40 participants per age group were included in the study. In the 1-way ANOVA, the NAMA, LAMA, and IMAT were significantly different between the age groups in both male and female participants (P≤.004), whereas the TAMA showed a significant difference only in male participants (male, P<.001; female, P=.88). The myosteatosis area had a strong negative correlation with muscle densities (r=-0.833 to -0.894), a moderate positive correlation with visceral fat areas (r=0.607 to 0.669), and a weak positive correlation with the subcutaneous fat areas (r=0.305 to 0.441). CONCLUSIONS: The automated web-based toolkit is feasible and enables quantitative CT assessment of myosteatosis, which can be a potential quantitative biomarker for evaluating structural and functional changes brought on by aging in the skeletal muscle.

Assessment of liver fibrosis severity using computed tomography-based liver and spleen volumetric indices in patients with chronic liver disease.

OBJECTIVES: To evaluate whether the liver and spleen volumetric indices, measured on portal venous phase CT images, could be used to assess liver fibrosis severity in chronic liver disease. METHODS: From 2007 to 2017, 558 patients (mean age 48.7 ± 13.1 years; 284 men and 274 women) with chronic liver disease (n = 513) or healthy liver (n = 45) were retrospectively enrolled. The liver volume (sVolL) and spleen volume (sVolS), normalized to body surface area and liver-to-spleen volume ratio (VolL/VolS), were measured on CT images using a deep learning algorithm. The correlation between the volumetric indices and the pathologic liver fibrosis stages combined with the presence of decompensation (F0, F1, F2, F3, F4C [compensated cirrhosis], and F4D [decompensated cirrhosis]) were assessed using Spearman's correlation coefficient. The performance of the volumetric indices in the diagnosis of advanced fibrosis, cirrhosis, and decompensated cirrhosis were evaluated using the area under the receiver operating characteristic curve (AUC). RESULTS: The sVolS (ρ = 0.47-0.73; p < .001) and VolL/VolS (ρ = -0.77-- 0.48; p < .001) showed significant correlation with liver fibrosis stage in all etiological subgroups (i.e., viral hepatitis, alcoholic and non-alcoholic fatty liver, and autoimmune diseases), while the significant correlation of sVolL was noted only in the viral hepatitis subgroup (ρ = - 0.55; p < .001). To diagnose advanced fibrosis, cirrhosis, and decompensated cirrhosis, the VolL/VolS (AUC 0.82-0.88) and sVolS (AUC 0.82-0.87) significantly outperformed the sVolL (AUC 0.63-0.72; p < .001). CONCLUSION: The VolL/VolS and sVolS may be used for assessing liver fibrosis severity in chronic liver disease. KEY POINTS: • Volumetric indices of liver and spleen measured on computed tomography images may allow liver fibrosis severity to be assessed in patients with chronic liver disease.

Influence of B1-Inhomogeneity on Pharmacokinetic Modeling of Dynamic Contrast-Enhanced MRI: A Simulation Study.

OBJECTIVE: To simulate the B1-inhomogeneity-induced variation of pharmacokinetic parameters on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). MATERIALS AND METHODS: B1-inhomogeneity-induced flip angle (FA) variation was estimated in a phantom study. Monte Carlo simulation was performed to assess the FA-deviation-induced measurement error of the pre-contrast R1, contrast-enhancement ratio, Gd-concentration, and two-compartment pharmacokinetic parameters (Ktrans, ve, and vp). RESULTS: B1-inhomogeneity resulted in -23-5% fluctuations (95% confidence interval [CI] of % error) of FA. The 95% CIs of FA-dependent % errors in the gray matter and blood were as follows: -16.7-61.8% and -16.7-61.8% for the pre-contrast R1, -1.0-0.3% and -5.2-1.3% for the contrast-enhancement ratio, and -14.2-58.1% and -14.1-57.8% for the Gd-concentration, respectively. These resulted in -43.1-48.4% error for Ktrans, -32.3-48.6% error for the ve, and -43.2-48.6% error for vp. The pre-contrast R1 was more vulnerable to FA error than the contrast-enhancement ratio, and was therefore a significant cause of the Gd-concentration error. For example, a -10% FA error led to a 23.6% deviation in the pre-contrast R1, -0.4% in the contrast-enhancement ratio, and 23.6% in the Gd-concentration. In a simulated condition with a 3% FA error in a target lesion and a -10% FA error in a feeding vessel, the % errors of the pharmacokinetic parameters were -23.7% for Ktrans, -23.7% for ve, and -23.7% for vp. CONCLUSION: Even a small degree of B1-inhomogeneity can cause a significant error in the measurement of pharmacokinetic parameters on DCE-MRI, while the vulnerability of the pre-contrast R1 calculations to FA deviations is a significant cause of the miscalculation.

Perfusion Assessment Using Intravoxel Incoherent Motion-Based Analysis of Diffusion-Weighted Magnetic Resonance Imaging: Validation Through Phantom Experiments.

OBJECTIVES: The aims of this study were to demonstrate the theoretical meaning of intravoxel incoherent motion (IVIM) parameters and to compare the robustness of 2 biexponential fitting methods through magnetic resonance experiments using IVIM phantoms. MATERIALS AND METHODS: Intravoxel incoherent motion imaging was performed on a 3 T magnetic resonance imaging scanner using 15 b values (0-800 s/mm) for 4 phantoms with different area fractions of the flowing water compartment (FWC%), at the infusion flow rates of 0, 1, 2, and 3 mL/min. Images were quantitatively analyzed using monoexponential free biexponential, and segmented biexponential fitting models. RESULTS: There were some inconsistent variations in Dslow with changing flow rates. The perfusion fraction, f, showed a significant positive correlation with the flow rate for both the free and segmented fitting methods (ρ = 0.838 to 0.969; P < 0.001). The fast diffusion coefficient, Dfast, had a significant positive correlation with the flow rate for segmented fitting (ρ = 0.745 to 0.969; P < 0.001), although it showed an inverse correlation with the flow rate for free fitting (ρ = -0.527 to -0.791; P ≤ 0.017). Significant positive correlations with the FWC% of the phantoms were noted for f (P = 0.510 for free fitting and P = 0.545 for segmented fitting, P < 0.001). CONCLUSIONS: The IVIM model allows for an approximate segmentation of molecular diffusion and perfusion, with a minor contribution of the perfusion effect on Dslow. The f and Dfast can provide a rough estimation of the flow fraction and flow velocity. Segmented fitting may be a more robust method than free fitting for calculating the IVIM parameters, especially for Dfast.