Leveraging radiomics and AI for precision diagnosis and prognostication of liver malignancies.

Liver tumors, whether primary or metastatic, have emerged as a growing concern with substantial global health implications. Timely identification and characterization of liver tumors are pivotal factors in order to provide optimum treatment. Imaging is a crucial part of the detection of liver tumors; however, conventional imaging has shortcomings in the proper characterization of these tumors which leads to the need for tissue biopsy. Artificial intelligence (AI) and radiomics have recently emerged as investigational opportunities with the potential to enhance the detection and characterization of liver lesions. These advancements offer opportunities for better diagnostic accuracy, prognostication, and thereby improving patient care. In particular, these techniques have the potential to predict the histopathology, genotype, and immunophenotype of tumors based on imaging data, hence providing guidance for personalized treatment of such tumors. In this review, we outline the progression and potential of AI in the field of liver oncology imaging, specifically emphasizing manual radiomic techniques and deep learning-based representations. We discuss how these tools can aid in clinical decision-making challenges. These challenges encompass a broad range of tasks, from prognosticating patient outcomes, differentiating benign treatment-related factors and actual disease progression, recognizing uncommon response patterns, and even predicting the genetic and molecular characteristics of the tumors. Lastly, we discuss the pitfalls, technical limitations and future direction of these AI-based techniques.

Approaches, advantages, and challenges to photon counting detector and multi-energy CT.

Photon counting detector CT (PCD-CT) is the newest major development in CT technology and has been commercially available since 2021. It offers major technological advantages over current standard-of-care energy integrating detector CT (EID-CT) including improved spatial resolution, improved iodine contrast to noise ratio, multi-energy imaging, and reduced noise. This article serves as a foundational basis to the technical approaches and concepts of PCD-CT technology with primary emphasis on detector technology in direct comparison to EID-CT. The article also addresses current technological challenges to PCD-CT with particular attention to cross talk and its causes (e.g., Compton scattering, fluorescence, charge sharing, K-escape) as well as pile-up.

Quantitative radiomics and qualitative LI-RADS imaging descriptors for non-invasive assessment of ß-catenin mutation status in hepatocellular carcinoma.

PURPOSE: Gain-of-function mutations in CTNNB1, gene encoding for ß-catenin, are observed in 25-30% of hepatocellular carcinomas (HCCs). Recent studies have shown ß-catenin activation to have distinct roles in HCC susceptibility to mTOR inhibitors and resistance to immunotherapy. Our goal was to develop and test a computational imaging-based model to non-invasively assess ß-catenin activation in HCC, since liver biopsies are often not done due to risk of complications. METHODS: This IRB-approved retrospective study included 134 subjects with pathologically proven HCC and available ß-catenin activation status, who also had either CT or MR imaging of the liver performed within 1 year of histological assessment. For qualitative descriptors, experienced radiologists assessed the presence of imaging features listed in LI-RADS v2018. For quantitative analysis, a single biopsy proven tumor underwent a 3D segmentation and radiomics features were extracted. We developed prediction models to assess the ß-catenin activation in HCC using both qualitative and quantitative descriptors. RESULTS: There were 41 cases (31%) with ß-catenin mutation and 93 cases (69%) without. The model's AUC was 0.70 (95% CI 0.60, 0.79) using radiomics features and 0.64 (0.52, 0.74; p = 0.468) using qualitative descriptors. However, when combined, the AUC increased to 0.88 (0.80, 0.92; p = 0.009). Among the LI-RADS descriptors, the presence of a nodule-in-nodule showed a significant association with ß-catenin mutations (p = 0.015). Additionally, 88 radiomics features exhibited a significant association (p < 0.05) with ß-catenin mutations. CONCLUSION: Combination of LI-RADS descriptors and CT/MRI-derived radiomics determine ß-catenin activation status in HCC with high confidence, making precision medicine a possibility.

Congestive Hepatopathy: Pathophysiology, Workup, and Imaging Findings with Pathologic Correlation.

Liver congestion is increasingly encountered in clinical practice and presents diagnostic pitfalls of which radiologists must be aware. The complex altered hemodynamics associated with liver congestion leads to diffuse parenchymal changes and the development of benign and malignant nodules. Distinguishing commonly encountered benign hypervascular lesions, such as focal nodular hyperplasia (FNH)-like nodules, from hepatocellular carcinoma (HCC) can be challenging due to overlapping imaging features. FNH-like lesions enhance during the hepatic arterial phase and remain isoenhancing relative to the background liver parenchyma but infrequently appear to wash out at delayed phase imaging, similar to what might be seen with HCC. Heterogeneity, presence of an enhancing capsule, washout during the portal venous phase, intermediate signal intensity at T2-weighted imaging, restricted diffusion, and lack of uptake at hepatobiliary phase imaging point toward the diagnosis of HCC, although these features are not sensitive individually. It is important to emphasize that the Liver Imaging Reporting and Data System (LI-RADS) algorithm cannot be applied in congested livers since major LI-RADS features lack specificity in distinguishing HCC from benign hypervascular lesions in this population. Also, the morphologic changes and increased liver stiffness caused by congestion make the imaging diagnosis of cirrhosis difficult. The authors discuss the complex liver macro- and microhemodynamics underlying liver congestion; propose a more inclusive approach to and conceptualization of liver congestion; describe the pathophysiology of liver congestion, hepatocellular injury, and the development of benign and malignant nodules; review the imaging findings and mimics of liver congestion and hypervascular lesions; and present a diagnostic algorithm for approaching hypervascular liver lesions. ©RSNA, 2024 Test Your Knowledge questions for this article are available in the supplemental material.

Detection of moderate hepatic steatosis on contrast-enhanced dual-source dual-energy CT: Role and accuracy of virtual non-contrast CT.

PURPOSE: To investigate diagnostic accuracy of virtual non contrast (VNC) images, based on dual-source dual-energy CT (dsDECT), for detection of at least moderate steatosis and to define a threshold value to make this diagnosis on VNC. METHODS: This single-institution retrospective study included patients who had multi-phasic protocol dsDECT. Regions of interests were placed in different segments of the liver and spleen on true non-contrast (TNC), VNC, and portal-venous phase (PVP) images. At least moderate steatosis was defined as liver attenuation (LHU) < 40 HU on TNC. Diagnostic performance of VNC to detect steatosis was determined and the new threshold was tested in a validation cohort. RESULTS: 236 patients were included in training cohort. Mean liver attenuation values were 51.3 ±â€¯10.8 HU and 58.1 ±â€¯11.5 HU for TNC and VNC (p < 0.001), with a mean difference (VNC - TNC) of 6.8 ±â€¯6.9 HU. Correlation between TNC and VNC was strong (r = 0.81, p < 0.001). The AUCs of LHU on VNC for detection of hepatic steatosis were 0.92 (95 % Cl: 0.86-0.98), 0.92 (95 % Cl: 0.87-0.97), 0.92 (95 % Cl: 0.86-0.99), 0.91 (95 % Cl: 0.84-0.97), and 0.87 (95 % Cl: 0.80-0.95) for entire liver, left lateral, left medial, right anterior, and right posterior segments, respectively. VNC had sensitivity/specificity of 100 % /42 % when using a threshold of 40 HU; they were 69 % and 95 %, respectively, when using optimized threshold of 46 HU. This threshold showed similar performance in validation cohort (n = 80). CONCLUSIONS: Hepatic attenuation on VNC has promising performance for detection of at least moderate steatosis. Proposed threshold of 46 HU provides high specificity and moderate sensitivity to detect steatosis.

Safety and feasibility of establishing an adjuvant hepatic artery infusion program.

BACKGROUND: Hepatic artery infusion (HAI) is less frequently used in the adjuvant setting for resectable colorectal liver metastasis (CRLM) due to concerns regarding toxicity. Our objective was to evaluate the safety and feasibility of establishing an adjuvant HAI program. METHODS: Patients who underwent HAI pump placement between January 2019 and February 2023 for CRLM were identified. Complications and HAI delivery were compared between patients who received HAI in the unresectable and adjuvant settings. RESULTS: Of 51 patients, 23 received HAI for unresectable CRLM and 28 in the adjuvant setting. Patients with unresectable CRLM more commonly had bilobar disease (n = 23/23 vs n = 18/28, p < 0.01) and more preoperative liver metastases (median 10 [IQR 6-15] vs 4 [IQR 3-7], p < 0.01). Biliary sclerosis was the most common complication (n = 2/23 vs n = 4/28); however, there were no differences in postoperative or HAI-specific complications. In the most recent two years, 0 patients in the unresectable group vs 2 patients in the adjuvant group developed biliary sclerosis. All patients were initiated on HAI with no difference in treatment times or dose reductions. CONCLUSION: Adjuvant HAI is safe and feasible for patients with resectable CRLM. HAI programs can carefully consider including patients with resectable CRLM if managed by an experienced multidisciplinary team with quality assurance controls in place.

Liver surface nodularity and ascites are associated with mortality risk in acute alcohol-associated hepatitis.

BACKGROUND: Acute alcohol-associated hepatitis (AH) is associated with high mortality. CT-derived liver surface nodularity (LSN) is a robust prognostic biomarker in other chronic liver diseases. The aim of this study was to determine relationships between LSN, disease severity, and mortality in AH. METHODS: Adults hospitalized with AH from January 2016 to March 2020 were included if an abdominal CT was performed between 8 weeks prior to 72 h after hospitalization. LSN was measured using quantitative methods (Liver Surface Nodularity Software version 0.88, Birmingham, AL, USA). Cox proportional hazards models, logistic regression and AUROC analysis were used to examine relationships between LSN and 180-day transplant-free survival. RESULTS: Of 386 patients hospitalized with AH during the study period, 230 had CT scans performed, and 205 met inclusion criteria. Mean transplant-free survival was 127 days (95% CI 118-137). Within each cohort, patients were grouped into low [LSN-LOW, N = 109 (53.2%)] and high [LSN-HIGH, N = 96 (46.8%)] LSN strata based on an optimal cutoff of 2.86 derived from unadjusted ROC curves. Patients with high LSN had features of portal hypertension, which included encephalopathy [53 (55.2%) vs. 43 (39.4%), p = 0.017], ascites on CT [81 (84.4%) vs. 69 (63.3%), p = 0.001] and portosystemic shunts [78 (81.2%) vs. 69 (63.3%), p = 0.003]. High LSN, ascites and MELD were independently associated with lower likelihood of 180-day transplant-free survival, and inclusion of a score assigning 1 point each for high LSN or ascites on CT (AHRADS score) to MELD enhanced diagnostic accuracy of AUROC for 180-day survival compared to MELD alone [AUROC 0.782 (95% CI 0.719-0.845) vs. 0.735 (0.667-0.802), p = 0.023]. CONCLUSIONS: CT-derived factors that include LSN and ascites are radiographic biomarkers associated with 180-day transplant-free survival in alcohol-associated hepatitis.

Imaging Evaluation of the Living Liver Donor: A Systems-Based Approach.

Liver transplant is the definitive treatment of end-stage liver disease and early hepatocellular carcinoma. The number of liver transplant surgeries done is highly affected by the number and availability of deceased donor organs. Living donor liver transplantation has emerged as an alternative source of donors, increasing the availability of organs for transplant. Many factors must be considered when choosing living donor candidates to maintain a high level of donor safety and organ survival. To that end, potential donors undergo a rigorous pre-donation workup.

Intestinal and Multivisceral Transplantation: Indications and Surgical Techniques.

Intestinal transplantation and multivisceral transplantation are technically challenging and complex procedures mainly performed on patients with irreversible and non-medically manageable end-stage intestinal failure. Increasingly, other organs besides small intestines are included in the allograft for which the terms "composite intestinal transplantation" and "multivisceral transplantation" are used. Commonly, complex vascular reconstructions are used for these procedures. Knowledge of surgical anatomy hence is essential for accurate interpretation of postoperative imaging in these patients. This article reviews the indications and most common surgical techniques for intestinal and multivisceral transplantations.

Intestinal and Multivisceral Transplantation: Complications.

Advancements in immunosuppression protocols, surgical techniques, and postoperative care in the last few decades have improved outcomes of intestinal transplant patients. Normal immediate postoperative imaging appearance can simulate pathology. Intestinal transplant recipients are prone for several postoperative complications due to the complex surgical technique, which involves multiple anastomoses, and immunogenic nature of the allograft intestine. Imaging plays a crucial role in detection of several major complications including infectious, immunologic, vascular, gastrointestinal, pancreaticobiliary, genitourinary, and neoplastic complications. The awareness of the posttransplant anatomy and normal imaging appearances helps radiologists anticipate and accurately detect posttransplant complications.

Utilizing CT to identify clinically significant biliary dilatation in symptomatic post-cholecystectomy patients: when should we be worried?

PURPOSE: To determine a reliable threshold common duct diameter on CT, in combination with other ancillary CT and clinical parameters, at which the likelihood of pathology requiring further imaging or intervention is increased in post-cholecystectomy patients. METHODS: In this IRB approved retrospective study, two attending radiologists independently reviewed CT imaging for 118 post-cholecystectomy patients, who were subsequently evaluated with MRCP, ERCP, or EUS, prompted by findings on the CT and clinical status. Measurements of the common duct (CD) were obtained at the porta hepatis, distal duct, and point of maximal dilation on axial and coronal CT scans. Patients were grouped by whether they required intervention after follow-up imaging. Pertinent baseline lab values and patient demographics were reviewed. RESULTS: Of the 118 post-cholecystectomy patients, 38 patients (31%) required intervention, and 80 patients (69%) did not require intervention after follow-up imaging. For both readers, axial and coronal CD diameters were significantly higher in the 'intervention required' vs 'no intervention' groups at all locations (p value < 0.05). There was good to excellent inter-reader agreement at all locations (ICC 0.68-0.92). Pertinent baseline lab values including AST (p = 0.043), ALT (p = 0.001), alkaline phosphatase (p = 0.0001), direct bilirubin (p = 0.011), total bilirubin (p = 0.028), and WBC (p = 0.043) were significantly higher in the 'intervention required' group. CD thresholds of 8 mm yielded the highest sensitivities (76-95%), and CD thresholds of 12 mm yielded the highest specificities (65-78%). CD combined with bilirubin levels increased sensitivity and specificity, compared to using either feature alone. CONCLUSION: Dilated CD on CT combined with bilirubin levels increases the sensitivity and specificity for identifying patients needing intervention. We recommend that a post-cholecystectomy patient who presents with a CD diameter > 10 mm on CT and elevated bilirubin levels should undergo further clinical and imaging follow-up.

Large Regenerative Nodules and Focal Nodular Hyperplasia-Like Lesions: Definition, Pathogenesis, and Imaging Findings.

Focal nodular hyperplasia-like (FNH-like) nodules are hepatocellular lesions with similar radiologic and pathologic features as typical FNH but occur within an abnormal liver. They arise due to alteration of hepatic vasculature at both the microscopic and macroscopic levels. Although these nodules are not thought to have malignant potential, their imaging features overlap with premalignant and malignant lesions including hepatocellular carcinoma (HCC) and arise in patients who may be at risk for HCC, posing a diagnostic and management dilemma. It is important to consider these benign entities when reviewing liver imaging of patients at risk for HCC to reduce unnecessary interventions.

Association of CT findings in patients with hemoperitoneum due to ruptured ovarian cysts with subsequent intervention.

PURPOSE: Hemoperitoneum caused by ovarian cyst rupture may necessitate intervention. The literature is lacking in descriptions of CT findings that help guide patient management. The purpose of this study is to consider CT findings associated with a need for intervention (surgical or interventional radiology management) versus conservative treatment (observation, pain management, and blood transfusions). METHODS: Two radiologists retrospectively and independently reviewed 103 CTs of pre-menopausal women who presented with acute hemoperitoneum related to ruptured ovarian cysts between January 2010 and January 2019. The following imaging features were assessed: ovarian cyst characteristics, sentinel clot, contrast extravasation, and hemoperitoneum size. Findings were correlated with patient demographics, clinical parameters, and management with surgery/interventional radiology procedure (intervention group) versus conservative management (conservative group). RESULTS: Of the 103 patients with hemoperitoneum from cyst rupture, 16% (n = 16) required intervention, and 84% underwent conservative treatment (n = 87). Length of stay (p = .008) was higher in the intervention group. Statistically significant CT findings in the intervention versus conservative group included (p-value reader 1/p-value reader 2) greatest AP dimension of hemoperitoneum (p = .001/p = 0.02), posterior cul-de-sac AP dimension (p = 0.03/p = .006), total cul-de-sac AP dimension (p = .002/p = .007), and number of spaces with hemoperitoneum (p = .01/p = .02). There was good to excellent inter-reader agreement for these findings (ICC 0.68-0.91). Active contrast extravasation was significant for one reader (p = .02) with poor inter-reader agreement (ICC 0.36). In utilizing ROC curves, thresholds of 107 mm (greatest axial AP dimension of hemoperitoneum) yielded a sensitivity and specificity of 0.81 and 0.62 for reader 1 and 0.69 and 0.55 for reader 2; 45 mm (posterior cul-de-sac AP dimension) yielded a sensitivity and specificity of 0.75 and 0.63 for reader 1 and 0.94 and 0.49 for reader 2; 70 mm (total cul-de-sac AP dimension) yielded a sensitivity and specificity of 0.75 and 0.64 for reader 1 and 0.75 and 0.50 for reader 2; and greater than 5 spaces yielded a sensitivity and specificity of 0.75 and 0.58 for reader 1 and 0.69 and 0.70 for reader 2. CONCLUSION: CT findings associated with intervention in hemoperitoneum due to ovarian cyst rupture include size of hemoperitoneum, number of abdominopelvic spaces with hemoperitoneum, and contrast extravasation.

Intra-patient comparison of 3D and 2D magnetic resonance elastography techniques for assessment of liver stiffness.

PURPOSE: To evaluate performance of 3D magnetic resonance elastography (MRE) using spin-echo echo-planar imaging (seEPI) for assessment of hepatic stiffness compared with 2D gradient-recalled echo (GRE) and 2D seEPI sequences. METHODS: Fifty-seven liver MRE examinations including 2D GRE, 2D seEPI, and 3D seEPI sequences were retrospectively evaluated. Elastograms were analyzed by 2 radiologists and polygonal regions of interests (ROIs) were drawn in 2 different fashions: "curated" ROI (avoiding liver edge, major vessels, and areas of wave interferences) and "non-curated" ROI (including largest cross section of liver, to assess the contribution of artifacts). Liver stiffness measurement (LSM) was calculated as the arithmetic mean of individual stiffness values for each technique. For 3D MRE, LSMs were also calculated based on 4 slices ("abbreviated LSM"). Intra-patient variations in LSMs and different methods of ROI placement were assessed by univariate tests. A p-value of < 0.05 was set as a statistically significant difference. RESULTS: Mean surface areas of the ROIs were 50,723 mm2, 12,669 mm2, 5814 mm2, and 10,642 mm2 for 3D MRE, abbreviated 3D MRE, 2D GRE, and 2D seEPI, respectively. 3D LSMs based on curated and non-curated ROIs showed no clinically significant difference, with a mean difference less than 0.1 kPa. Abbreviated 3D LSMs had excellent correlation with 3D LSMs based on all slices (r = 0.9; p < 0.001) and were not significantly different (p = 0.927). CONCLUSION: 3D MRE allows more reproducible measurements due to its lower susceptibility to artifacts and provides larger areas of parenchyma, enabling a more comprehensive evaluation of the liver.

Pancreatic Cystic Lesions and Malignancy: Assessment, Guidelines, and the Field Defect.

The widespread use of high-spatial-resolution cross-sectional imaging has led to an increase in detection of incidental pancreatic cystic lesions. These lesions are a diverse group, ranging from indolent and premalignant lesions to invasive cancers. The diagnosis of several of these lesions can be suggested on the basis of their imaging appearance, while many other lesions require follow-up imaging and/or aspiration. The smaller cystic lesions, often branch-duct intraductal papillary mucinous neoplasms, have overlapping imaging characteristics that make diagnostic assessment of the natural history and malignancy risk confusing. Expert panels have developed societal guidelines, based on a consensus, for surveillance of these lesions. However, these guidelines are often inconsistent and are constantly evolving as additional scientific data are accumulated. Identification of features associated with increased risk of malignancy is important for proper management. The concept of field defect, whereby pancreatic adenocarcinoma develops at a site different from the site of the pancreatic cyst, adds to the complexity of screening guidelines. As a result of the differences in guidelines, key stakeholders (eg, radiologists, gastroenterologists, and surgeons) must review and come to a consensus regarding which guideline, or combination of guidelines, to follow at their individual institutions. Standardized reporting and macros are helpful for ensuring the uniformity of interpretations. Radiologists play a critical role in the detection and characterization of pancreatic cystic lesions, in the follow-up recommendations for these lesions, and in the detection of associated cancer. An invited commentary by Zaheer is available online. Online supplemental material is available for this article. ©RSNA, 2021.

Magnetic Resonance Elastography of the Liver: A Single-Institution Review of Cases Performed in a 5-Year Interval.

OBJECTIVE: This study aimed to assess the diagnostic accuracy of magnetic resonance elastography (MRE) in detecting hepatic fibrosis and determining clinically relevant stiffness cutoff values per stage of fibrosis. METHODS: This retrospective study assessed 1488 hepatic MRE evaluations performed at a single institution for 5 years. Mean liver stiffness measurements were collected from 282 patients who had an MRE study within 1 year of histopathologic analysis. Areas under receiver operating characteristic curves were calculated for each stage of fibrosis with nonparametric ordinal measures of accuracy, and Youden Index was determined. RESULTS: Mean liver stiffness measurement values were as follows: F0, 2.5± 0.55 kPa; F1, 3.1± 0.80 kPa; F2, 3.4±0.95 kPa; F3, 4.7±1.44 kPa; and F4, 7.9± 2.64 kPa. Nonparametric ordinal measures of accuracy per fibrosis stage were as follows: F0: 0.934, P < 0.001; F0-F1: 0.917, P < 0.001; F0-F2: 0.944, P < 0.001; and F0-F3: 0.941, P < 0.001. Youden Index values for fibrosis stages F2, F3, and F4 were 3.9, 4.0, and 4.5 kPa, respectively. CONCLUSIONS: Magnetic resonance elastography is an accurate diagnostic tool in assessing liver fibrosis.

Transformer based Generative Adversarial Network for Liver Segmentation.

Automated liver segmentation from radiology scans (CT, MRI) can improve surgery and therapy planning and follow-up assessment in addition to conventional use for diagnosis and prognosis. Although convolutional neural networks (CNNs) have became the standard image segmentation tasks, more recently this has started to change towards Transformers based architectures because Transformers are taking advantage of capturing long range dependence modeling capability in signals, so called attention mechanism. In this study, we propose a new segmentation approach using a hybrid approach combining the Transformer(s) with the Generative Adversarial Network (GAN) approach. The premise behind this choice is that the self-attention mechanism of the Transformers allows the network to aggregate the high dimensional feature and provide global information modeling. This mechanism provides better segmentation performance compared with traditional methods. Furthermore, we encode this generator into the GAN based architecture so that the discriminator network in the GAN can classify the credibility of the generated segmentation masks compared with the real masks coming from human (expert) annotations. This allows us to extract the high dimensional topology information in the mask for biomedical image segmentation and provide more reliable segmentation results. Our model achieved a high dice coefficient of 0.9433, recall of 0.9515, and precision of 0.9376 and outperformed other Transformer based approaches. The implementation details of the proposed architecture can be found at https://github.com/UgurDemir/tranformer_liver_segmentation.