Systematic Analysis of Common Factors Impacting Deep Learning Model Generalizability in Liver Segmentation.

Purpose: To investigate the effect of training data type on generalizability of deep learning liver segmentation models. Materials and Methods: This Health Insurance Portability and Accountability Act-compliant retrospective study included 860 MRI and CT abdominal scans obtained between February 2013 and March 2018 and 210 volumes from public datasets. Five single-source models were trained on 100 scans each of T1-weighted fat-suppressed portal venous (dynportal), T1-weighted fat-suppressed precontrast (dynpre), proton density opposed-phase (opposed), single-shot fast spin-echo (ssfse), and T1-weighted non-fat-suppressed (t1nfs) sequence types. A sixth multisource (DeepAll) model was trained on 100 scans consisting of 20 randomly selected scans from each of the five source domains. All models were tested against 18 target domains from unseen vendors, MRI types, and modality (CT). The Dice-Sørensen coefficient (DSC) was used to quantify similarity between manual and model segmentations. Results: Single-source model performance did not degrade significantly against unseen vendor data. Models trained on T1-weighted dynamic data generally performed well on other T1-weighted dynamic data (DSC = 0.848 ± 0.183 [SD]). The opposed model generalized moderately well to all unseen MRI types (DSC = 0.703 ± 0.229). The ssfse model failed to generalize well to any other MRI type (DSC = 0.089 ± 0.153). Dynamic and opposed models generalized moderately well to CT data (DSC = 0.744 ± 0.206), whereas other single-source models performed poorly (DSC = 0.181 ± 0.192). The DeepAll model generalized well across vendor, modality, and MRI type and against externally sourced data. Conclusion: Domain shift in liver segmentation appears to be tied to variations in soft-tissue contrast and can be effectively bridged with diversification of soft-tissue representation in training data.Keywords: Convolutional Neural Network (CNN), Deep Learning Algorithms, Machine Learning Algorithms, Supervised Learning, CT, MRI, Liver Segmentation Supplemental material is available for this article. © RSNA, 2023.

A Faster Prostate MRI: Comparing a Novel Denoised, Single-Average T2 Sequence to the Conventional Multiaverage T2 Sequence Regarding Lesion Detection and PI-RADS Score Assessment.

BACKGROUND: The T2 w sequence is a standard component of a prostate MRI examination; however, it is time-consuming, requiring multiple signal averages to achieve acceptable image quality. PURPOSE/HYPOTHESIS: To determine whether a denoised, single-average T2 sequence (T2 -R) is noninferior to the standard multiaverage T2 sequence (T2 -S) in terms of lesion detection and PI-RADS score assessment. STUDY TYPE: Retrospective. POPULATION: A total of 45 males (age range 60-75 years) who underwent clinically indicated prostate MRI examinations, 21 of whom had pathologically proven prostate cancer. FIELD STRENGTH/SEQUENCE: A 3 T; T2 w FSE, DWI with ADC maps, and dynamic contrast-enhanced images with color-coded perfusion maps. T2 -R images were created from the raw data utilizing a single "average" with iterative denoising. ASSESSMENT: Nine readers randomly assessed complete exams including T2 -R and T2 -S images in separate sessions. PI-RADS version 2.1 was used. All readers then compared the T2 -R and T2 -S images side by side to evaluate subjective preference. An additional detailed image quality assessment was performed by three senior level readers. STATISTICAL TESTS: Generalized linear mixed effects models for differences in lesion detection, image quality features, and overall preference between T2 -R and T2 -S sequences. Intraclass correlation coefficients (ICC) were used to assess reader agreement for all comparisons. A significance threshold of P = 0.05 was used for all statistical tests. RESULTS: There was no significant difference between sequences regarding identification of lesions with PI-RADS ≥3 (P = 0.10) or PI-RADS score (P = 0.77). Reader agreement was excellent for lesion identification (ICC = 0.84). There was no significant overall preference between the two sequences regarding image quality (P = 0.07, 95% CI: [-0.23, 0.01]). Reader agreement was good regarding sequence preference (ICC = 0.62). DATA CONCLUSION: Use of single-average, denoised T2 -weighted images was noninferior in prostate lesion detection or PI-RADS scoring when compared to standard multiaverage T2 -weighted images. EVIDENCE LEVEL: 3. TECHNICAL EFFICACY: Stage 3.

CT-derived body composition measurements as predictors for neoadjuvant treatment tolerance and survival in gastroesophageal adenocarcinoma.

PURPOSE: Treatment for gastroesophageal adenocarcinomas can result in significant morbidity and mortality. The purpose of this study is to supplement methods for choosing treatment strategy by assessing the relationship between CT-derived body composition, patient, and tumor features, and clinical outcomes in this population. METHODS: Patients with neoadjuvant treatment, biopsy-proven gastroesophageal adenocarcinoma, and initial staging CTs were retrospectively identified from institutional clinic encounters between 2000 and 2019. Details about patient, disease, treatment, and outcomes (including therapy tolerance and survival) were extracted from electronic medical records. A deep learning semantic segmentation algorithm was utilized to measure cross-sectional areas of skeletal muscle (SM), visceral fat (VF), and subcutaneous fat (SF) at the L3 vertebra level on staging CTs. Univariate and multivariate analyses were performed to assess the relationships between predictors and outcomes. RESULTS: 142 patients were evaluated. Median survival was 52 months. Univariate and multivariate analysis showed significant associations between treatment tolerance and SM and VF area, SM to fat and VF to SF ratios, and skeletal muscle index (SMI) (p = 0.004-0.04). Increased survival was associated with increased body mass index (BMI) (p = 0.01) and increased SMI (p = 0.004). A multivariate Cox model consisting of BMI, SMI, age, gender, and stage demonstrated that patients in the high-risk group had significantly lower survival (HR = 1.77, 95% CI = 1.13-2.78, p = 0.008). CONCLUSION: CT-based measures of body composition in patients with gastroesophageal adenocarcinoma may be independent predictors of treatment complications and survival and can supplement methods for assessing functional status during treatment planning.

Data-Driven Modification of the LI-RADS Major Feature System on Gadoxetate Disodium-Enhanced MRI: Toward Better Sensitivity and Simplicity.

BACKGROUND: The Liver Imaging Reporting and Data System (LI-RADS) is widely accepted as a reliable diagnostic scheme for hepatocellular carcinoma (HCC) in at-risk patients. However, its application is hampered by substantial complexity and suboptimal diagnostic sensitivity. PURPOSE: To propose data-driven modifications to the LI-RADS version 2018 (v2018) major feature system (rLI-RADS) on gadoxetate disodium (EOB)-enhanced magnetic resonance imaging (MRI) to improve sensitivity and simplicity while maintaining high positive predictive value (PPV) for detecting HCC. STUDY TYPE: Retrospective. POPULATION: Two hundred and twenty-four consecutive at-risk patients (training dataset: 169, independent testing dataset: 55) with 742 LR-3 to LR-5 liver observations (HCC: N = 498 [67%]) were analyzed from a prospective observational registry collected between July 2015 and September 2018. FIELD STRENGTH/SEQUENCE: 3.0 T/T2-weighted fast spin-echo, diffusion-weighted spin-echo based echo-planar and three-dimensional (3D) T1-weighted gradient echo sequences. ASSESSMENT: All images were evaluated by three independent abdominal radiologists who were blinded to all clinical, pathological, and follow-up information. Composite reference standards of either histopathology or imaging follow-up were used. STATISTICAL TESTS: In the training dataset, LI-RADS v2018 major features were used to develop rLI-RADS based on their associated PPV for HCC. In an independent testing set, diagnostic performances of LI-RADS v2018 and rLI-RADS were computed using a generalized estimating equation model and compared with McNemar's test. A P value <0.05 was considered statistically significant. RESULTS: The median (interquartile range) size of liver observations was 13 mm (7-27 mm). The diagnostic table for rLI-RADS encompassed 9 cells, as opposed to 16 cells for LI-RADS v2018. In the testing set, compared to LI-RADS v2018, rLI-RADS category 5 demonstrated a significantly superior sensitivity (76% vs. 61%) while maintaining comparably high PPV (92.5% vs. 94.1%, P = 0.126). DATA CONCLUSION: Compared with LI-RADS v2018, rLI-RADS demonstrated improved simplicity and significantly superior diagnostic sensitivity for HCC in at-risk patients. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY STAGE: 2.

How frequently does hepatocellular carcinoma develop in at-risk patients with a negative liver MRI examination with intravenous Gadobenate dimeglumine?

OBJECTIVE: To determine the rate of development of clinically significant liver nodules (LR-4, LR-5, LR-M) after a negative MRI in an HCC screening population. METHODS: This retrospective study included patients at risk of developing HCC requiring imaging surveillance who had undergone multiphase Gadobenate dimeglumine-enhanced MRI that was negative and had follow up LI-RADS compliant multiphase CTs or MRIs for at least 12 months or positive follow-up within 12 months. Follow-up examinations were classified as negative (no nodules or only LR-1 nodules) or positive (nodule other than LR-1). Time-to-first positive examination, types of nodules, and cumulative incidence of nodule development were recorded. RESULTS: 204 patients (mean age 58.9 ± 10.2 years, 128 women), including 172 with cirrhosis, were included. Based CT/MRI follow-up (median 35 months, range 12-80 months), the overall cumulative incidence of developing a nodule was 10.5%. Cumulative incidence of nodule development was: 0.5% at 6-9 months and 2.1% at 12 ± 3 months, including one LR-4 nodule, one LR-M nodule, and two LR-3 nodules. The cumulative incidence of clinically significant nodule development was 1.1% at 9-15 months. 70% (143/204) of patients also underwent at least one US follow-up, and no patient developed a positive US examination following index negative MRI. CONCLUSION: Clinically significant liver nodules develop in 1.1% of at-risk patients in the first year following negative MRI. While ongoing surveillance is necessary for at-risk patients, our study suggests than longer surveillance intervals after a negative MRI may be reasonable and that further research is needed to explore this possibility.

Relative Sarcopenia With Excess Adiposity Predicts Survival After Transjugular Intrahepatic Portosystemic Shunt Creation.

OBJECTIVE. The purpose of this study was to assess the impact of relative sarcopenia with excess adiposity on mortality after transjugular intrahepatic portosystemic shunt (TIPS) creation. MATERIALS AND METHODS. In this single-institution retrospective study, patients underwent abdominal CT scans within 100 days before or 30 days after TIPS creation. Subcutaneous and visceral adipose tissue and muscle were segmented at the L3 vertebral level. Relative sarcopenia with excess adiposity was defined as the lowest sex-specific quartile of muscle area divided by muscle plus adipose. Dates of death, liver transplantation, TIPS occlusion, and hepatic encephalopathy (HE) after TIPS creation were identified. Mortality was evaluated using competing risks survival analysis. Number of HE episodes and time to first episode were analyzed using negative binomial regression and competing risks survival analysis, respectively. RESULTS. A total of 141 patients (91 men; mean age, 56 years) were included in this study. In univariate analyses, Model for End-Stage Liver Disease (MELD) score (hazard ratio [HR], 1.09 per point; CI, 1.05-1.13; p < 0.001) and relative sarcopenia with excess adiposity (HR, 2.70; CI, 1.55-4.69; p < 0.001) were significant risk factors for shorter survival after TIPS. In multivariate analysis, both MELD score (HR, 1.09; CI, 1.03-1.15; p = 0.003) and relative sarcopenia with excess adiposity (HR, 2.65; CI, 1.56-4.51; p < 0.001) were significant predictors of worse survival. The C-index at 30 days was 0.71 for MELD score, 0.72 for relative sarcopenia with excess adiposity, and 0.80 for a model including both. There was no association between relative sarcopenia with excess adiposity and number of HE episodes (incidence rate ratio, 1.08; CI, 0.49-2.40; p = 0.84) or time to first HE episode (HR, 0.89; CI, 0.51-1.54; p = 0.67). CONCLUSION. Relative sarcopenia with excess adiposity is a risk factor for mortality after TIPS and contributes additional prognostic information beyond MELD score.

Multicenter Safety and Practice for Off-Label Diagnostic Use of Ferumoxytol in MRI.

Background Ferumoxytol is approved for use in the treatment of iron deficiency anemia, but it can serve as an alternative to gadolinium-based contrast agents. On the basis of postmarketing surveillance data, the Food and Drug Administration issued a black box warning regarding the risks of rare but serious acute hypersensitivity reactions during fast high-dose injection (510 mg iron in 17 seconds) for therapeutic use. Whereas single-center safety data for diagnostic use have been positive, multicenter data are lacking. Purpose To report multicenter safety data for off-label diagnostic ferumoxytol use. Materials and Methods The multicenter ferumoxytol MRI registry was established as an open-label nonrandomized surveillance databank without industry involvement. Each center monitored all ferumoxytol administrations, classified adverse events (AEs) using the National Cancer Institute Common Terminology Criteria for Adverse Events (grade 1-5), and assessed the relationship of AEs to ferumoxytol administration. AEs related to or possibly related to ferumoxytol injection were considered adverse reactions. The core laboratory adjudicated the AEs and classified them with the American College of Radiology (ACR) classification. Analysis of variance was used to compare vital signs. Results Between January 2003 and October 2018, 3215 patients (median age, 58 years; range, 1 day to 96 years; 1897 male patients) received 4240 ferumoxytol injections for MRI. Ferumoxytol dose ranged from 1 to 11 mg per kilogram of body weight (≤510 mg iron; rate ≤45 mg iron/sec). There were no systematic changes in vital signs after ferumoxytol administration (P > .05). No severe, life-threatening, or fatal AEs occurred. Eighty-three (1.9%) of 4240 AEs were related or possibly related to ferumoxytol infusions (75 mild [1.8%], eight moderate [0.2%]). Thirty-one AEs were classified as allergiclike reactions using ACR criteria but were consistent with minor infusion reactions observed with parenteral iron. Conclusion Diagnostic ferumoxytol use was well tolerated, associated with no serious adverse events, and implicated in few adverse reactions. Registry results indicate a positive safety profile for ferumoxytol use in MRI. © RSNA, 2019 Online supplemental material is available for this article.

Acute Acalculous Cholecystitis in Hospitalized Patients With Hematologic Malignancies and Prognostic Importance of Gallbladder Ultrasound Findings.

OBJECTIVES: Patients with hematologic malignancies, especially those with acute disease or those receiving intense chemotherapy, are known to develop acute acalculous cholecystitis (AAC). The aim of this study was to evaluate the diagnostic and prognostic value of the established ultrasound (US) diagnostic criteria for AAC in patients with acute hematologic malignancies who were clinically suspected to have AAC. METHODS: We retrospectively studied the US findings of the gallbladder in patients with hematologic malignancies and correlated these findings with the duration of clinical symptoms, complications, and gallbladder-specific mortality. The major criteria were a 3.5-mm or thicker wall, pericholecystic fluid, intramural gas, and a sloughed mucosal membrane. The minor criteria were echogenic bile and hydrops (gallbladder distension > 4 cm). Ultrasound findings were considered positive if they included 2 major criteria or 1 major and 2 minor criteria. RESULTS: Ninety-four (25.5%) of 368 patients with hematologic malignancies had clinical signs of AAC during their acute phase of illness or during intense chemotherapy. Forty-three (45.7%) of these 94 patients had AAC-positive test results based on US criteria. The mean duration of symptoms was significantly longer (7.8 days) in this group than among the patients with negative test results (3.9 days; P < .001). Patients with positive test results had a higher rate of complications or mortality (20.9%) than those with negative test results (0%; P < .001). CONCLUSIONS: Symptomatic patients who meet the US criteria for the diagnosis of AAC have a poor prognosis. Other patients require a close follow-up US examination within 1 week to detect early progression.