Search

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

The Project Baseline Health Study (PBHS) was launched to map human health through a comprehensive understanding of both the health of an individual and how it relates to the broader population. The study will contribute to the creation of a biomedical information system that accounts for the highly complex interplay of biological, behavioral, environmental, and social systems. The PBHS is a prospective, multicenter, longitudinal cohort study that aims to enroll thousands of participants with diverse backgrounds who are representative of the entire health spectrum. Enrolled participants will be evaluated serially using clinical, molecular, imaging, sensor, self-reported, behavioral, psychological, environmental, and other health-related measurements. An initial deeply phenotyped cohort will inform the development of a large, expanded virtual cohort. The PBHS will contribute to precision health and medicine by integrating state of the art testing, longitudinal monitoring and participant engagement, and by contributing to the development of an improved platform for data sharing and analysis.

Microvascular invasion (MVI) is a critical risk factor for postoperative recurrence of hepatocellular carcinoma (HCC). Preknowledge of MVI would assist tailored surgery planning in HCC management. In this multicenter study, we aimed to explore the validity of deep learning (DL) in MVI prediction using two imaging modalities—contrast-enhanced computed tomography (CE-CT) and gadoxetic acid-enhanced magnetic resonance imaging (EOB-MRI). A total of 750 HCCs were enrolled from five Chinese tertiary hospitals. Retrospective CE-CT (n = 306, collected between March, 2013 and July, 2019) and EOB-MRI (n = 329, collected between March, 2012 and March, 2019) data were used to train two DL models, respectively. Prospective external validation (n = 115, collected between July, 2015 and February, 2018) was performed to assess the developed models. Furthermore, DL-based attention maps were utilized to visualize high-risk MVI regions. Our findings revealed that the EOB-MRI-based DL model achieved superior prediction outcome to the CE-CT-based DL model (area under receiver operating characteristics curve (AUC): 0.812 vs. 0.736, p = 0.038, sensitivity: 70.4% vs. 57.4%, p = 0.015, specificity: 80.3% vs. 86.9%, p = 0.052). DL attention maps could visualize peritumoral high-risk areas with genuine histopathologic confirmation. Both DL models could stratify high and low-risk groups regarding progression free survival and overall survival (p &lt, 0.05). Thus, DL can be an efficient tool for MVI prediction, and EOB-MRI was proven to be the modality with advantage for MVI assessment than CE-CT.

BACKGROUND: Proton density fat fraction (PDFF) estimated by using chemical shift–encoded (CSE) MRI is an accepted imaging biomarker of hepatic steatosis. This work aims to promote standardized use of CSE MRI to estimate PDFF. PURPOSE: To assess the accuracy of CSE MRI methods for estimating PDFF by determining the linearity and range of bias observed in a phantom. MATERIALS AND METHODS: In this prospective study, a commercial phantom with 12 vials of known PDFF values were shipped across nine U.S. centers. The phantom underwent 160 independent MRI examinations on 27 1.5-T and 3.0-T systems from three vendors. Two three-dimensional CSE MRI protocols with minimal T1 bias were included: vendor and standardized. Each vendor’s confounder-corrected complex or hybrid magnitude-complex based reconstruction algorithm was used to generate PDFF maps in both protocols. The Siemens reconstruction required a configuration change to correct for water-fat swaps in the phantom. The MRI PDFF values were compared with the known PDFF values by using linear regression with mixed-effects modeling. The 95% CIs were calculated for the regression slope (ie, proportional bias) and intercept (ie, constant bias) and compared with the null hypothesis (slope = 1, intercept = 0). RESULTS: Pooled regression slope for estimated PDFF values versus phantom-derived reference PDFF values was 0.97 (95% CI: 0.96, 0.98) in the biologically relevant 0%–47.5% PDFF range. The corresponding pooled intercept was −0.27% (95% CI: −0.50%, −0.05%). Across vendors, slope ranges were 0.86–1.02 (vendor protocols) and 0.97–1.0 (standardized protocol) at 1.5 T and 0.91–1.01 (vendor protocols) and 0.87–1.01 (standardized protocol) at 3.0 T. The intercept ranges (absolute PDFF percentage) were −0.65% to 0.18% (vendor protocols) and −0.69% to −0.17% (standardized protocol) at 1.5 T and −0.48% to 0.10% (vendor protocols) and −0.78% to −0.21% (standardized protocol) at 3.0 T. CONCLUSION: Proton density fat fraction estimation derived from three-dimensional chemical shift–encoded MRI in a commercial phantom was accurate across vendors, imaging centers, and field strengths, with use of the vendors’ product acquisition and reconstruction software. © RSNA, 2021 See also the editorial by Dyke in this issue.

The primary objective was to compare the performance of 3 different abbreviated MRI (AMRI) sets extracted from a complete gadoxetate-enhanced MRI obtained for hepatocellular carcinoma (HCC) screening. Secondary objective was to perform a preliminary cost-effectiveness analysis, comparing each AMRI set to published ultrasound performance for HCC screening in the USA. This retrospective study included 237 consecutive patients (M/F, 146/91; mean age, 58 years) with chronic liver disease who underwent a complete gadoxetate-enhanced MRI for HCC screening in 2017 in a single institution. Two radiologists independently reviewed 3 AMRI sets extracted from the complete exam: non-contrast (NC-AMRI: T2-weighted imaging (T2wi)+diffusion-weighted imaging (DWI)), dynamic-AMRI (Dyn-AMRI: T2wi+DWI+dynamic T1wi), and hepatobiliary phase AMRI (HBP-AMRI: T2wi+DWI+T1wi during the HBP). Each patient was classified as HCC-positive/HCC-negative based on the reference standard, which consisted in all available patient data. Diagnostic performance for HCC detection was compared between sets. Estimated set characteristics, including historical ultrasound data, were incorporated into a microsimulation model for cost-effectiveness analysis. The reference standard identified 13/237 patients with HCC (prevalence, 5.5%; mean size, 33.7 ± 30 mm). Pooled sensitivities were 61.5% for NC-AMRI (95% confidence intervals, 34.4–83%), 84.6% for Dyn-AMRI (60.8–95.1%), and 80.8% for HBP-AMRI (53.6–93.9%), without difference between sets (p range, 0.06–0.16). Pooled specificities were 95.5% (92.4–97.4%), 99.8% (98.4–100%), and 94.9% (91.6–96.9%), respectively, with a significant difference between Dyn-AMRI and the other sets (p

Khaled M Elsayes,1 Ania Z Kielar,2 Michelle M Agrons,3 Janio Szklaruk,1 An Tang,4 Mustafa R Bashir,5 Donald G Mitchell,6 Richard K Do,7 Kathryn J Fowler,8 Victoria Chernyak,9 Claude B Sirlin10 1Department of Diagnostic Radiology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA; 2Department of Diagnostic Radiology, University of Ottawa, Ottawa, ON, Canada; 3Department of Diagnostic Radiology, Baylor College of Medicine, Houston, TX, USA; 4Department of Radiology, Radio-Oncology and Nuclear Medicine, Université de Montréal, Montreal, QC, Canada; 5Department of Diagnostic Radiology, Duke University School of Medicine, Durham, NC, 6Department of Diagnostic Radiology, Thomas Jefferson University, Philadelphia, PA, 7Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, 8Mallinckrodt Institute of Radiology, Washington University in Saint Louis, Saint Louis, MO, 9Department of Radiology Albert Einstein College of Medicine, Bronx, New York, NY, 10Department of Diagnostic Radiology, University of California, San Diego, CA,USA Abstract: The increasing incidence and high morbidity and mortality of hepatocellular carcinoma (HCC) have inspired the creation of the Liver Imaging Reporting and Data System (LI-RADS). LI-RADS aims to reduce variability in exam interpretation, improve communication, facilitate clinical therapeutic decisions, reduce omission of pertinent information, and facilitate the monitoring of outcomes. LI-RADS is a dynamic process, which is updated frequently. In this article, we describe the LI-RADS 2014 version (v2014), which marks the second update since the initial version in 2011. Keywords: hepatocellular carcinoma, imaging, reporting, cirrhosis, hyperenhancement washout

Aim: The aim of this study was to evaluate deformable image registration (DIR)-based motion estimation of the liver for four-dimensional-computed tomography (4D-CT) and 4D-magnetic resonance imaging (MRI). Methods: Five liver cancer patients were included. Each patient was imaged with 4D-CT and 4D-MRI under an Institutional Review Board-approved protocol. Motion estimation of the liver was obtained by performing DIR on 4D-CT and 4D-MRI. A region of interest (ROI) encompassing the expert-determined gross tumor volume was used as surrogate to evaluate the accuracy of the motion estimation. ROI motion trajectories were estimated by averaging the displacement vector fields (DVFs) within the ROI during the breathing cycles for 4D-CT and 4D-MRI and were compared to those extracted from cine MR. Target registration error (TRE), correlation coefficient (CC) for phase agreement, difference in phase at maximum displacement (ΔPmax), and Dice's Similarity Coefficient (DSC) for overall motion agreement were determined. Results: As compared to 4D-CT, 4D-MRI resulted in smaller TRE in DVFs (anterior-posterior [AP]: 1.0 ± 0.4 mm vs. 1.5 ± 0.5 mm, superior-inferior [SI]: 1.9 ± 0.7 mm vs. 2.2 ± 0.8 mm), greater CC (AP: 0.67 ± 0.32 vs. 0.49 ± 0.26, SI: 0.84 ± 0.15 vs. 0.58 ± 0.28), smaller ΔPmax (AP: 1.4 ± 1.7 vs. 2.0 ± 1.0, SI: 0.4 ± 0.9 vs. 1.2 ± 0.8), and greater DSC (AP: 0.67 ± 0.08 vs. 0.61 ± 0.11, SI: 0.73 ± 0.12 vs. 0.67 ± 0.10). Conclusion: 4D-MRI can potentially provide more realistic respiratory DVFs of the liver than 4D-CT.

N-acetylcysteine (NAC) has been well studied in the treatment of acetaminophen-induced and select non-acetaminophen-induced liver failure. However, its role in the management of sickle cell hepatic crisis resulting in acute liver failure (ALF) is unknown. We describe and discuss the novel and beneficial use of NAC in a 25-year-old man with ALF due to sickle cell hepatic crisis. We further review ALF in sickle cell disease and NAC in the treatment of non-acetaminophen-induced liver failure. Our case highlights the promising role of NAC in sickle cell-related liver injury.

OBJECTIVE. The purpose of this study was to determine the diagnostic performance of hepatobiliary phase hypointensity and Liver Imaging Reporting and Data System (LI-RADS) major imaging features in the diagnosis of hepatocellular carcinoma (HCC) in hepatic lesions with arterial phase hyperenhancement (APHE) measuring 10-19 mm in patients at high risk of HCC. MATERIALS AND METHODS. A composite reference standard of pathologic analysis and imaging follow-up was used. The diagnostic performance (sensitivity and specificity) of hepatobiliary phase hypointensity and LI-RADS major imaging features other than APHE for the diagnosis of HCC was assessed and compared by means of a logistic regression model. RESULTS. This retrospective dual-institution study included 189 LI-RADS LR-3 and LR-4 lesions measuring 10-19 mm and having APHE in 144 consecutively registered patients (96 men, 48 women; mean age, 58 years). Hepatobiliary phase hypointensity had significantly higher sensitivity (84% [92/109], p < 0.00001) than major imaging features in the diagnosis of HCC but lower specificity (84% [67/80]; p = 0.01). However, hepatobiliary phase hypointensity in LR-3 observations measuring 10-19 mm and having APHE had moderately elevated sensitivity (73% [44/60]) and specificity (85%, 64/75). All three major imaging features had high specificity for the diagnosis of HCC, including 95% (76/80) for washout, 100% (80/80) for enhancing capsule, and 99% (79/80) for threshold growth. CONCLUSION. Major imaging features have high specificity for the diagnosis of HCC in lesions measuring 10-19 mm that have APHE. The finding of hepatobiliary phase hypointensity significantly improves sensitivity while moderately high specificity is maintained for the diagnosis of HCC in LR-3 lesions measuring 10-19 mm that exhibit APHE.

© 2018, Springer Science+Business Media, LLC, part of Springer Nature. The Liver Imaging and Reporting Data System (LI-RADS) is a comprehensive system for standardizing the terminology, technique, interpretation, reporting, and data collection of liver imaging with the overarching goal of improving communication, clinical care, education, and research relating to patients at risk for or diagnosed with hepatocellular carcinoma (HCC). In 2018, the American Association for the Study of Liver Diseases (AASLD) integrated LI-RADS into its clinical practice guidance for the imaging-based diagnosis of HCC. The harmonization between the AASLD and LI-RADS diagnostic imaging criteria required minor modifications to the recently released LI-RADS v2017 guidelines, necessitating a LI-RADS v2018 update. This article provides an overview of the key changes included in LI-RADS v2018 as well as a look at the LI-RADS v2018 diagnostic algorithm and criteria, technical recommendations, and management suggestions. Substantive changes in LI-RADS v2018 are the removal of the requirement for visibility on antecedent surveillance ultrasound for LI-RADS 5 (LR-5) categorization of 10-19 mm observations with nonrim arterial phase hyper-enhancement and nonperipheral “washout”, and adoption of the Organ Procurement and Transplantation Network definition of threshold growth (≥ 50% size increase of a mass in ≤ 6 months). Nomenclatural changes in LI-RADS v2018 are the removal of -us and -g as LR-5 qualifiers.

This article provides a glimpse into the future of the Liver Imaging Reporting and Data System (LI-RADS), discussing the immediate and long-term plans for its continuing improvement and expansion. To complement the Core and Essentials components of the latest version of LI-RADS, a comprehensive manual will be released soon, and it will include technical recommendations, management guidance, as well as reporting instructions and templates. In this article, we briefly review the process by which LI-RADS has been developed until now, a process guided by a variable combination of data, expert opinion, and desire for congruency with other diagnostic systems in North America. We then look forward, envisioning that forthcoming updates to LI-RADS will occur regularly every 3 to 5 years, driven by emerging high-quality scientific evidence. We highlight some of the key knowledge and technology gaps that will need to be addressed to enable the needed refinements. We also anticipate future expansions in scope to meet currently unaddressed clinical needs. Finally, we articulate a vision for eventual unification of imaging system for HCC screening and surveillance, diagnosis and staging, and treatment response assessment.

Proper pre- and post-transplant diagnostic imaging work-up is fundamental in ensuring a successful outcome for renal transplantation. Despite exposure to ionizing radiation, CT has high spatial resolution and is a widely available and fast imaging technique. CT is performed routinely to delineate the anatomy of the kidney, relevant vasculature, and urinary collecting system in the living donor, to assess the iliac vessels in potential recipients prior to surgery, and to assess early and late-term post-transplant complications. The purpose of this article is to outline the optimal CT protocol and the main reportable findings for both the donor and the recipient diagnostic imaging work-up as well as to point out the main issues regarding ionizing radiation exposure and contrast medium injection in these subjects.

The Liver Imaging Reporting and Data System (LI-RADS) standardizes the interpretation, reporting, and data collection for imaging examinations in patients at risk for hepatocellular carcinoma (HCC). It assigns category codes reflecting relative probability of HCC to imaging-detected liver observations based on major and ancillary imaging features. LI-RADS also includes imaging features suggesting malignancy other than HCC. Supported and endorsed by the American College of Radiology (ACR), the system has been developed by a committee of radiologists, hepatologists, pathologists, surgeons, lexicon experts, and ACR staff, with input from the American Association for the Study of Liver Diseases and the Organ Procurement Transplantation Network/United Network for Organ Sharing. Development of LI-RADS has been based on literature review, expert opinion, rounds of testing and iteration, and feedback from users. This article summarizes and assesses the quality of evidence supporting each LI-RADS major feature for diagnosis of HCC, as well as of the LI-RADS imaging features suggesting malignancy other than HCC. Based on the evidence, recommendations are provided for or against their continued inclusion in LI-RADS. (©) RSNA, 2017 Online supplemental material is available for this article.

Purpose We report an approach, termed Repeated k-t-subsampling and artifact-minimization (ReKAM), for removing motion artifacts in free-breathing abdominal MRI. The method is particularly valuable for challenging patients who may not hold their breath for a long time or have irregular respiratory rate. Methods The ReKAM framework comprises one acquisition module and two reconstruction modules. A fast MRI sequence is used to repeatedly acquire multiple sets of k-t space data. Motion artifacts are then minimized by two reconstruction modules: 1) a bootstrapping module in k-t-space is used to identify a low-artifact image; 2) a constrained reconstruction module that integrates projection onto convex set (POCS) and multiplexed sensitivity encoding (MUSE), termed POCSMUSE, is applied to further remove residual artifact. The ReKAM framework is compatible with different pulse sequences, and generally applicable to irregular data sampling patterns in k-space. Free-breathing fast spin-echo MRI data, acquired from healthy volunteers and patients, were used to evaluate the developed ReKAM method. Results Experimental results show that the ReKAM technique can produce high-quality free-breathing images with the artifact levels comparable to that of breath-holding MRI. Conclusion The ReKAM framework improves the quality of free-breathing abdominal MRI data, and is compatible with various MRI pulse sequences. This article is protected by copyright. All rights reserved.

To assess the impact of patient habitus, acquisition parameters, detector efficiencies, and reconstruction techniques on the accuracy of iodine quantification using dual-source dual-energy CT (DECT).Two phantoms simulating small and large patients contained 20 iodine solutions mimicking vascular and parenchymal enhancement from saline isodensity to 400 HU and 30 iodine solutions simulating enhancement of the urinary collecting system from 400 to 2,000 HU. DECT acquisition (80/140 kVp and 100/140 kVp) was performed using two DECT systems equipped with standard and integrated electronics detector technologies. DECT raw datasets were reconstructed using filtered backprojection (FBP), and iterative reconstruction (SAFIRE I/V).Accuracy for iodine quantification was significantly higher for the small compared to the large phantoms (9.2 % ± 7.5 vs. 24.3 % ± 26.1, P = 0.0001), the integrated compared to the conventional detectors (14.8 % ± 20.6 vs. 18.8 % ± 20.4, respectively; P = 0.006), and SAFIRE V compared to SAFIRE I and FBP reconstructions (15.2 % ± 18.1 vs. 16.1 % ± 17.6 and 18.9 % ± 20.4, respectively; P ≤ 0.003). A significant synergism was observed when the most effective detector and reconstruction techniques were combined with habitus-adapted dual-energy pairs.In a second-generation dual-source DECT system, the accuracy of iodine quantification can be substantially improved by an optimal choice and combination of acquisition parameters, detector, and reconstruction techniques.

BACKGROUND AND PURPOSE: There are no guidelines for reporting incidental thyroid nodules seen on CT and MR imaging. We evaluated radiologists9 current reporting practices for incidental thyroid nodules detected on these imaging modalities. MATERIALS AND METHODS: Radiologists were surveyed regarding their reporting practices by using 14 scenarios of incidental thyroid nodules differing in size, patient demographics, and clinical history. Scenarios were evaluated for the following: 1) radiologists9 most commonly selected response, and 2) the proportion of radiologists selecting that response (degree of agreement). These measures were used to determine how the patient scenario and characteristics of the radiologists affected variability in practice. RESULTS: One hundred fifty-three radiologists participated. In 8/14 scenarios, the most common response was to “recommend sonography.” For the other scenarios, the most common response was to “report in only body of report.” The overall mean agreement for the 14 scenarios was 53%, and agreement ranged from 36% to 75%. Smaller nodules had lower agreement: 43%–51% for 8-mm nodules compared with 64%–75% for 15-mm nodules. Agreement was poorest for the 10-mm nodule in a 60-year-old woman (36%) and for scenarios with additional history of lung cancer (39%) and multiple nodules (36%). There was no significant difference in reporting practices and agreement when radiologists were categorized by years of practice, practice type, and subspecialty (P > .55). CONCLUSIONS: The reporting practice for incidental thyroid nodules on CT or MR imaging is highly variable among radiologists, especially for patients with smaller nodules (≤10 mm) and patients with multiple nodules and a history of cancer. This variability highlights the need for practice guidelines.

Background To evaluate the contrast agent performance of Gd-EOB-DTPA and Gd-BOPTA for detection and assessment of extrahepatic findings, semi-quantitatively and qualitatively. Methods 13 patients with 19 extrahepatic lesions underwent liver MRI with Gd-EOB-DTPA and Gd-BOPTA. Quantitative and relative SNR measurements were performed in each dataset in the arterial and portalvenous phase within the extrahepatic lesion, aorta, inferior vena cava, portal vein, spleen, pancreas and renal cortex. Further, relative CNR measurements were performed. Three readers assessed contrast quality using a five-point scale and choosing the preferred image dataset. Statistical analysis consisted of a Student’s t-test with p

OBJECTIVE The aim of this study was to compare the image quality of abdominal computed tomography scans in an anthropomorphic phantom acquired at different radiation dose levels where each raw data set is reconstructed with both a standard convolution filtered back projection (FBP) and a full model-based iterative reconstruction (MBIR) algorithm. MATERIALS AND METHODS An anthropomorphic phantom in 3 sizes was used with a custom-built liver insert simulating late hepatic arterial enhancement and containing hypervascular liver lesions of various sizes. Imaging was performed on a 64-section multidetector-row computed tomography scanner (Discovery CT750 HD; GE Healthcare, Waukesha, WI) at 3 different tube voltages for each patient size and 5 incrementally decreasing tube current-time products for each tube voltage. Quantitative analysis consisted of contrast-to-noise ratio calculations and image noise assessment. Qualitative image analysis was performed by 3 independent radiologists rating subjective image quality and lesion conspicuity. RESULTS Contrast-to-noise ratio was significantly higher and mean image noise was significantly lower on MBIR images than on FBP images in all patient sizes, at all tube voltage settings, and all radiation dose levels (P < 0.05). Overall image quality and lesion conspicuity were rated higher for MBIR images compared with FBP images at all radiation dose levels. Image quality and lesion conspicuity on 25% to 50% dose MBIR images were rated equal to full-dose FBP images. CONCLUSION This phantom study suggests that depending on patient size, clinically acceptable image quality of the liver in the late hepatic arterial phase can be achieved with MBIR at approximately 50% lower radiation dose compared with FBP.