RPS 701 - Quantitative imaging and biomarkers

Author Block: C. Maino, P. N. Franco, C. R. G. L. O. M. Talei Franzesi, E. De Bernardi, L. Cristoferi, D. Ippolito; Monza/IT
Purpose: Risk assessment in primary sclerosing cholangitis (PSC) by magnetic resonance imaging (MRI) relies on semi‐quantitative analysis, which can result in interpretation variability. Radiomics may offer a quantitative approach for risk stratification. This study aims to explore and validate MRI‐derived radiomic features to identify high‐risk PSC patients.
Methods or Background: In this prospective study (January 2019–December 2022), consecutive PSC patients undergoing routine gadoxetate disodium‐enhanced MRI were recruited. Using PyRadiomics, whole liver parenchyma features were extracted from five MRI sequences according to the Image Biomarker Standardisation Initiative (IBSI). Patients were categorised into risk groups based on the Mayo risk score (MRS) and liver stiffness measurement (LSM). Features associated with high‐risk patients were selected and validated in an independent cohort. A survival analysis was conducted in the combined cohort to assess the prognostic value of the radiomic features for clinical events.
Results or Findings: One hundred and two PSC patients were enrolled in this study. Five radiomics features were associated with high risk in the training cohort. In the validation setting, GLRLM‐Run Entropy in the fat‐saturation T2 weighted imaging (FS‐T2W) sequence was the only significant feature, with an odds ratio of 3.90 (CI 1.46–10.42, p = 0.007) for MRS and 2.97 (CI 1.33–6.66, p = 0.008) for LSM. Its prognostic potential on clinical outcome was confirmed by Cox regression analysis in the combined cohort (hazard ratio per 0.1 increase = 1.480, CI 1.226–1.786), showing excellent predictive performance (C‐index = 0.857).
Conclusion: GLRLM‐Run Entropy in FS‐T2W is a novel radiomics‐based biomarker for risk stratification in PSC patients. It is quantitative, standardised, easy to compute and cost‐free, positioning it as a potential key innovation in PSC radiology‐based biomarkers
Limitations: Retrospective
Single institution
Funding for this study: None/not used
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approved

Author Block: L. Sannikova, S. Gusarova, I. Gruzdev, N. Karelskaya, V. Shirokov, V. Aznaurov, A. Ustalov, S. A. Shmeleva, E. V. Kondratyev; Moscow/RU
Purpose: To investigate how reproducible liver radiomics features are when derived from monoenergetic (MonoE) CT reconstructions, using portal venous phase imaging as the reference standard with a consistent 3D-ROI approach.
Methods or Background: The study included 64 abdominal CT examinations. A spherical VOI was positioned within a homogeneous segment of liver parenchyma devoid of vessels. This VOI, initially delineated in the portal venous phase, was automatically propagated across MonoE reconstructions (40, 55, 70, and 200 keV). Radiomic features were extracted and normalized via z-score transformation. Reproducibility was quantified using concordance correlation coefficient, intra-class correlation coefficient, and paired statistical tests with FDR adjustment.
Results or Findings: Reconstructions at 40 and 55 keV showed the closest alignment with portal imaging, yielding CCC values above 0.90 for most features, with ~45% qualifying as reproducible. The 70 keV level demonstrated intermediate reproducibility, especially limited for texture-based features, with only ~30% classified as reproducible. The 200 keV reconstructions performed poorly, with a high proportion of features below CCC 0.70, and evidence of systematic bias. Across all reconstructions, shape and first-order features remained the most robust. Similarity analyses and PCA revealed clustering of portal with MonoE40 and MonoE55, while MonoE200 consistently diverged from other groups.
Conclusion: Virtual monoenergetic reconstructions at 40 and 55 keV demonstrate the highest reproducibility of hepatic radiomics features and can be regarded as reliable surrogates for portal venous phase imaging. Importantly, these findings indicate that spectral CT data can be integrated alongside conventional radiomics models. Substantial energy shifts (e.g., high-keV reconstructions) generate significant alterations in feature behavior. Such investigations may not only refine technical reproducibility but also reveal hidden imaging phenotypes or disease-specific characteristics that remain undetectable at standard energy levels.
Limitations: CT-scanner from only one vendor (Philips) was used.
Funding for this study: Nothing to disclosure
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the local ethics committee of the A. V. Vishnevsky National Medical Research Center of Surgery.

Author Block: L. Jia¹, S. N. Li¹, Y. Li², J. Li¹, J. Lei¹; ¹Lanzhou/CN, ²Beijing/CN
Purpose: Vessels encapsulating tumor clusters (VETC) represent a distinct histologic vascular pattern strongly associated with metastatic spread in hepatocellular carcinoma (HCC). This study aimed to develop and validate a CT-based radiomics model for the preoperative prediction of VETC positivity using a dual-center design.
Methods or Background: In this retrospective study, we included patients with pathologically confirmed HCC who underwent contrast-enhanced CT between January 2019 and January 2023. The training cohort consisted of 142 patients from Center 1, and an independent external validation cohort included 61 patients from Center 2. Radiomic features were extracted from arterial and portal venous phase images. A radiomic signature was developed using a K-nearest neighbor classifier and compared to clinical-only and combined clinical-radiomic models. Performance was assessed using AUC, calibration curves, and decision curve analysis (DCA).
Results or Findings: The radiomics model significantly outperformed the clinical model, with an AUC of 0.894 (95% CI: 0.863–0.916) in the training set and 0.85 (95% CI: 0.78–0.92) in the external validation set. There was no significant difference between the radiomics and combined clinical-radiomics models (AUC: 0.922 vs. 0.894; p = 0.182). DCA demonstrated robust clinical utility of the radiomics model.
Conclusion: The proposed CT radiomics model noninvasively predicts VETC pattern in HCC with strong generalizability across institutions. This tool shows promise for identifying aggressive HCC subtypes preoperatively, potentially supporting personalized treatment strategies.
Limitations: None
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study was approved by the Institutional Review Board (LDYYLL-2024-398)

Author Block: J. CUI, Y. Lei, C. Ma, Y. Xie, Z. Luo; Jiangmen Central Hospital/CN
Purpose: To develop and validate a machine learning (ML) model using preoperative CT-derived hepatic vascular features for predicting 2-year recurrence-free survival (RFS) in hepatocellular carcinoma (HCC) patients after hepatectomy.
Methods or Background: This multicenter retrospective study enrolled 361 patients from two centers, allocated into training (n=181), internal validation (n=120), and external test (n=60) cohorts. From preoperative arterial phase and portal venous phase CT, we extracted 3D vascular features including volumes of the hepatic artery, portal vein, and hepatic vein, their vessel-to-liver volume ratios, and 3D fractal dimension as a measure of vascular complexity. We built three ML models to predict 2-year recurrence risk: a vascular model, a clinical model, and an integrated model combining both. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). Kaplan-Meier and log-rank tests assessed RFS differences between model-stratified risk groups.
Results or Findings: The integrated model demonstrated superior predictive performance for 2-year RFS, achieving the highest AUC values of 0.8057 in the internal validation set and 0.7736 in the external test set. It significantly outperformed the vascular model (AUCs: 0.7572 and 0.6900) and the clinical model (AUCs: 0.5848 and 0.6146). Furthermore, patients stratified as high-risk by the integrated model exhibited significantly worse actual RFS outcomes than those in the low-risk group (all P < 0.05).
Conclusion: Our study validates that the combined ML model, integrating 3D vascular features and clinical factors, serves as an effective non-invasive tool for predicting the risk of recurrence within 2 years after hepatectomy in HCC patients. Its superior performance over single-source models underscores the synergistic value of vascular complexity for improved prognostic stratification.
Limitations: This study is limited by its retrospective nature and requires larger, prospective, multi-center cohorts for further validation to ensure generalizability.
Funding for this study: No funding was received for this study
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: W. C. Bartholomä, S. Cai, C. Simonsson, M. Karlsson, S. Kechagias, M. Woisetschlager, N. Dahlström, P. Lundberg; Linköping/SE
Purpose: Primary sclerosing cholangitis (PSC) is a rare fibroinflammatory liver disease with a highly variable course, ranging from indolent to rapidly progressive cases leading to cirrhosis, liver failure, or hepatobiliary malignancies. Accurate identification of patients at risk for poor outcomes remains challenging. MRI-based approaches such as DiStrict, Anali, and relative enhancement show promise but are limited by operator dependency or static measurements. Pharmacokinetic modeling offers a dynamic, quantitative alternative for risk assessment in PSC.
Methods or Background: A prospective cohort of 26 PSC patients underwent up to four annual MRI exams with a total clinical follow-up of 7.5 years. Clinical endpoints included liver transplantation, decompensated cirrhosis, and cholangiocarcinoma. The pharmacokinetic model's performance was compared to MELD, the Amsterdam-Oxford Model (AOM), Relative Liver Enhancement (RE) and the Anali Score. Correlation and ROC analyses with laboratory values and established risk scores were performed. A segmental approach was also explored.
Results or Findings: The pharmacokinetic model (Globalki ) showed significant correlations with MELD (r = -0.429, p = 0.029), AOM (r = -0.557, p = 0.003), and endpoint events (r = -0.605, p = 0.001). ROC analysis indicated that Globalki (AUC 0.943) outperformed the Anali Scores (AUC 0.800 – 0.829) in identifying patients at risk for an endpoint event and was comparable to MELD (AUC 0.857) and AOM (AUC 0.900). The model showed no significant correlation with laboratory values, suggesting its potential as an independent biomarker.
Conclusion: The pharmacokinetic model based on hepatobiliary contrast uptake in MRI provides relevant independent prognostic information in PSC, comparable to established models such as AOM or MELD.
Limitations: Small sample size and limited number of endpoint events. Mayo Risk score could not be calculated for the patient cohort as a complement to the Amsterdam-Oxford Model.
Funding for this study: RFoU medel för doktorander (Regional Research and Development Funding for PhD students, provided by the Region of Östergötland (Östergötland Medical County). Additional funding from Vinnova (Sweden's Innovation Agency), and the Swedish Research Council (Engineering and Natural Sciences (VR/NT)).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approved by Institutional Review board, Dnr 2011/8-24-21/2, 2018/1111-32, and 2018/1494-31/3

Author Block: A. Fedak¹, T. J. Popiela², M. Zwolińska-Wcisło², K. Mirowski², J. Jamroś²; ¹Kraków/PL, ²Krakow/PL
Purpose: Background: Contrast-enhanced ultrasound (CEUS) is widely used for focal liver lesions, but its role in autoimmune liver diseases (AILD: autoimmune hepatitis [AIH], primary biliary cholangitis [PBC], and overlap) is unclear. Core biopsy is the gold standard but invasive. Non-invasive tools, including shear-wave elastography (SWE) and biochemical indices (e.g., FIB-4), are increasingly applied yet may not fully reflect microvascular changes. In AILD, lobular pathology alters resistance and promotes shunts. CEUS, by measuring portal-to-hepatic vein transit time (PV/HV TT), may capture these alterations.
Methods or Background: Methods: We prospectively studied 25 patients (22–66 years, 68% female) with histologically confirmed AIH (n=19), PBC (n=4), or overlap (n=2). All underwent core liver biopsy as the reference standard. Exclusion criteria were decompensated cirrhosis and coagulopathy. Patients undergoing CEUS for unrelated indications served as controls. The protocol included laboratory testing, FIB-4, Doppler ultrasound, SWE (Metavir), and CEUS with parametric imaging. Correlations with histology were analyzed using Spearman’s rank.
Results or Findings: Results: Advanced fibrosis by FIB-4 was found in 48% of patients; SWE indicated significant fibrosis (Metavir ≥F3) in 62%. A threshold of PV/HV TT <3.5 s emerged as the most accurate discriminator of lobular injury, yielding 92% sensitivity versus 72% for laboratory tests (p=0.18). In controls, mean PV/HV TT was 7.8 s. No patient showed concordant negativity across CEUS and laboratory testing. PV/HV TT negatively correlated with FIB-4 (ρ=–0.42, p=0.035). All imaging and laboratory indices correlated significantly with biopsy.
Conclusion: Conclusion: This pilot study shows that CEUS sensitively detects microvascular alterations in AILD and complements SWE and biochemical indices. With further validation, CEUS may reduce reliance on biopsy and improve patient monitoring
Limitations: As a pilot investigation the limitation of the study is small amount of participants.
Funding for this study: Jagiellonian University grant No 1072.6120.268.2022
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Obrazowanie ultrasonograficzne unaczynienia wątroby z wykorzystaniem środków kontrastujących w chorobach przebiegających z włóknieniem : niealkoholowa choroba stłuszczeniowa wątroby (NAFLD) autoimmunologiczne zapalenie wątroby (AIH), pierwotne zapalenie dróg żółciowych (PBC)” 1072.6120.268.2022, Acceptance of Bioethic Commitee of Jagiellonian University Medical College

Author Block: J. Gotta, L. D. Grünewald, V. Koch, S. Mahmoudi, T. Vogl; Frankfurt/DE
Purpose: Single-slice analysis at the third lumbar vertebra (L3) is widely used in oncologic imaging to assess body composition. However, three-dimensional volumetric analysis may better reflect tissue distribution. This study compared the prognostic performance of volumetric and L3-derived body composition metrics in patients with cholangiocarcinoma.
Methods or Background: In this retrospective study, 147 patients with histologically confirmed intrahepatic (n = 65), perihilar (n = 46), or distal (n = 36) cholangiocarcinoma (47 women, 100 men; median age 67.0 years, IQR 58.0–73.0) underwent native-phase CT. Body composition was analysed using nnU-Net-based segmentation with manual quality control. Volumetric parameters included muscle percentage, myosteatosis percentage, visceral adipose tissue (VAT) percentage, and total adipose tissue (TAT) percentage. L3-derived metrics comprised skeletal muscle area (SMA_L3), muscle radiation attenuation (SMRA_L3), intramuscular fat (IMAT_L3), VAT_L3, and VAT HU_L3. Prognostic value was assessed using Kaplan–Meier and Cox proportional hazards models.
Results or Findings: Volumetric muscle percentage and myosteatosis percentage were the strongest survival predictors (log-rank = 17.43 and 17.67; HR = 0.93 and 1.06; both p < 0.001), followed by VAT and TAT percentages (log-rank = 12.26 and 8.20; both p = 0.01). These volumetric metrics showed strong statistical significance and clear survival stratification. Among L3-derived metrics, IMAT_L3 (HR = 1.09; p = 0.002), SMRA_L3 (HR = 0.97; p =0.005), VAT HU_L3 (HR = 0.98; p =0.004), and SMA_L3 (HR = 0.99; p = 0.02) were associated with survival but with lower log-rank values and less consistent group separation. VAT_L3 and SAT_L3 were not significant.
Conclusion: Volumetric CT-derived body composition metrics showed stronger and more consistent associations with survival than corresponding L3-based measures. Despite representing similar compartments, volumetric analysis may better capture the heterogeneity and extent of tissue distribution, improving prognostic accuracy in cholangiocarcinoma.
Limitations: Retrospective Study
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Waiver due to retrospective nature

Author Block: V. Koch¹, L. D. Grünewald¹, J. Gotta¹, S. Mahmoudi¹, R. Hammerstingl¹, K. Eichler¹, O. Darwish², T. Vogl¹, R. Sinkus²; ¹Frankfurt/DE, ²Paris/FR
Purpose: The rise in metabolic syndrome is driving diffuse liver disease and HCC, often without advanced fibrosis. Identifying at-risk individuals in the pre-fibrotic stage is challenging. This study developed and validated a 3D MRE–based biomechanical imaging framework for early liver disease detection and characterization of HCC-related alterations in non-fibrotic livers.
Methods or Background: In this prospective study, 3D MRE was consecutively performed in 193 participants (mean age, 56 years; 83 women) with suspected or confirmed liver disease, including 73 biopsy-confirmed cases (38%) and 26 HCCs (14%). 20 healthy volunteers served as controls. The complex shear modulus was decomposed into classical stiffness (|G*|), shear wave attenuation (α), and phase angle (Y), yielding a two-dimensional “biomechanical fingerprint” of liver tissue. Biomechanical parameters were correlated with histopathology, blood biomarkers, and clinical characteristics.
Results or Findings: We identified three biomechanical regions: healthy tissue, a pre-fibrotic niche with inflammation but near-normal stiffness, and fibrotic/cirrhotic tissue. In the pre-fibrotic niche, α was markedly reduced despite preserved stiffness, accompanied by decreased E-cadherin and early α-smooth muscle actin upregulation. An α threshold <50 m⁻¹ was linked to higher HCC risk, even without fibrosis, and the biomechanical continuum mirrored gradual changes in clinical and molecular biomarkers (FIB-4, ASAT, PDFF, platelet count).
Conclusion: Our findings show that 3D MRE allows organ-specific, noninvasive mapping of liver disease and identifies a biomechanical state permissive to HCC before fibrosis. This approach extends elastography beyond fibrosis staging and enables early, individualized HCC risk stratification.
Limitations: This study has limitations, including predominance of fatty liver disease, reliance on biopsy with inherent variability, reduced biopsy numbers due to MRI eligibility, and its observational, hypothesis-generating design. Technical constraints (single frequency, single vendor, research transducer) and reliance on presumed F0 in most healthy volunteers may limit reproducibility.
Funding for this study: This work was supported by the German Doktor Robert Pfleger Foundation, the Wilhelm Vaillant Foundation, the French Integrated Cancer Research Center “SiRIC InsiTu” (INCa-DGOS-INSERM-ITMO Cancer_18008), and the French National Research Agency RHU Operandi (ANR-21-RHUS-0012).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The institutional ethics review board approved this prospective, cross-sectional, single-center study that complies with the Declaration of Helsinki.

Author Block: L. Huimin, X. Wu, Z. Xu, F. Yan; Shanghai/CN
Purpose: Dual-energy CT-derived hepatic extracellular volume (ECV) has shown promise in liver fibrosis quantification but is limited by the higher radiation dose of dual-energy acquisitions and the need for specific scan protocols. Photon-counting detector CT (PCCT), representing a new generation of multienergy CT, offers intrinsic spectral images and improved dose efficiency. This study aimed to investigate the feasibility of photon-counting CT-based ECV for liver fibrosis evaluation, using MR elastography (MRE) as the reference standard.
Methods or Background: Between June 2024 and July 2025, 53 participants were prospectively recruited to undergo multiphase abdominal PCCT and MRE. Delayed-phase PCCT images were acquired after the portal venous phase for ECV quantification. Correlation between ECV and MRE-derived liver stiffness measurements (LSMs) was assessed in a head-to-head comparison, and stratification performance for different fibrosis stages was evaluated with receiver operating characteristic curve analysis.
Results or Findings: According to established MRE thresholds (3.50 kPa, 4.30 kPa, 6.50 kPa) for liver fibrosis staging, 11 participants had F1, 12 had F2, 20 had F3, and 10 had cirrhosis. ECV correlated strongly with LSM (Spearman ρ = 0.903 [95% CI: 0.834, 0.944], P < .0001). For fibrosis stratification, ECV demonstrated excellent diagnostic performance for ≥ F2, ≥ F3, and F4, with areas under the receiver operating characteristic curve of 0.96 (0.91-1.00), 0.93 (0.86-0.99), and 0.97 (0.94-1.00), respectively. The corresponding cutoff values were 23.35%, 24.15%, and 27.28%.
Conclusion: These findings demonstrated excellent correlation between ECV and LSM, supporting PCCT-derived ECV as a feasible alternative for noninvasive liver fibrosis quantification.
Limitations: This study was limited by its relatively small sample size from a single center, which may restrict the generalizability of the findings. Larger multicenter studies are warranted to validate the performance of PCCT-derived ECV in liver fibrosis evaluation.
Funding for this study: No funding was provided for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The ethics committee notification can be found under the number KY2024-187.