RPS 201 - What's new in biliary diseases?

Author Block: A. Grigoriadis¹, S. G. Hamma¹, G. Kemmerich², J. S. Nayagam³, K. Horsthuis⁴, M-C. Londoño⁵, D. Assis⁶, S. Charanjeet⁷, A. Bergquist¹; ¹Stockholm/SE, ²Oslo/NO, ³London/UK, ⁴Amsterdam/NL, ⁵Barcelona/ES, ⁶Connecticut, CT/US, ⁷New Haven, CT/US
Purpose: To validate the reproducibility and prognostic value of the DiStrict-score in a multicenter international cohort.
Methods or Background: The DiStrict-score is an MRCP-based classification of the severity of ductal changes (ranging from 1 to 8) and is reproducible and associated with liver-transplantation and liver-related death. However, it has not been validated. For this retrospective multicenter study with participation of eight international PSC-centers, hepatologists from each center retrieved data of consecutive adult PSC individuals (MRCP, demographics, liver-tests, PSC diagnosis date, hepatobiliary cancer development, liver- transplantation, death, and cause of death). Two radiologists from each center applied the DiStrict-score independently to the patients of their center. Cases of disagreement were resolved in consensus. Interreader agreement was assessed for each pair of radiologists with the intraclass correlation coefficient (ICC), with a two-way random-effects model, absolute-agreement, and single-measurement. The association of the DiStrict-score with outcomes (transplant-free survival, development of hepatobiliary malignancy) was assessed with Cox-regression. Survival estimates were calculated with Kaplan-Meier curves and the curves were compared with the log-rank test.
Results or Findings: 415 patients (271 males, 248 with ulcerative colitis) with median diagnosis age of 39 years, were included. During a median follow-up of 84 months 101 patients developed outcomes (liver-transplantation; n=78, liver-related death; n=10, hepatobiliary cancer; n=13). The interreader agreement ranged between 0.61 and 0.91 for the different centers. Patients with high DiStrict-scores (5–8) had a higher risk of developing outcomes compared to patients with low scores (1–4) (log-rank test; p=0.0008) with a hazard ratio of 1,98 (95%CI; 1.32-2.96, p=0.001).
Conclusion: The DiStrict-sore is reproducible with good interreader agreement and is associated with transplant-free survival and development of hepatobiliary malignancy.
Limitations: The limitations of the study are its retrospective design and the non-standardized MRCP acquisition technique.
Funding for this study: Funding was provided by Medical Diagnostics Karolinska.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Each center obtained ethical approval by local ethical committees .

Author Block: A. Grigoriadis, S. G. Hamma, A. Bergquist; Stockholm/SE
Purpose: To assess the prevalence and reproducibility of the evaluation of the presence of high-grade strictures (HGS) in MRCP, in PSC individuals. Moreover, to assess the predictive value of HGS for development of hepatobiliary malignancy, liver-transplantation, and liver-related death.
Methods or Background: AASLD and EASL have introduced in their guidelines for PSC the term HGS defined as a stricture seen in MRCP with >75% reduction of the lumen of common and/or biliary ducts. However, the prevalence, reproducibility of their detection, and their potential value for predicting outcomes have not been assessed. Two radiologists independently assessed the presence of HGS in MRCPs of 203 individuals with PSC recruited at Karolinska University Hospital in the SUPRIM study between 2012 and 2015. MRCP, demographic, clinical-laboratory and outcome data (hepatobiliary malignancy, liver-transplantation, and liver-related death) were retrieved for all patients. Interreader agreement of the evaluation of HGS was calculated with the intraclass correlation coefficient (ICC) using a two-way random-effects model, single-measurement, and absolute-agreement. The association of HGS with outcomes was assessed with Cox-regression. Survival estimates were calculated with Kaplan-Meier curves and the curves were compared with the log-rank test.
Results or Findings: After exclusion, 171 patients (103 males, 95 with ulcerative colitis) with a median diagnosis age of 40 years were included. During a median follow-up of 124 months 49 patients developed outcomes (liver-transplantation=36, liver-related death=5, hepatobiliary malignancy=8). 80 patients (47%) had HGS. The agreement was moderate with ICC=0.72 (95%CI; 0.64-0.78). Patients with HGS had a higher risk to develop outcomes (p=0.01) with a hazard-ratio of 2.08 (95%CI; 1.17-3.71).
Conclusion: HGS are common, can be identified with acceptable reproducibility, and are associated with outcomes.
Limitations: The limitations of the study are its retrospective design and that no intrareader agreement analysis was performed.
Funding for this study: Funding was provided by Medical Diagnostics Karolinska.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the Swedish ethical review authority (2011/824-31/2, 2018/1111-32, 2018/1494-31/3).

Author Block: X. Zhou¹, J. Hu², S-T. Feng¹; ¹Guangzhou/CN, ²Beijing/CN
Purpose: To preoperatively predict the Perineural invasion (PNI) in intrahepatic cholangiocarcinoma (ICC) on gadoxetic acid (EOB)-enhanced MRI, the deep learning with clinical model based fusion model was developed and evaluated.
Methods or Background: A total of 165 patients with pathologically diagnosed ICC who underwent preoperative EOB-enhanced MRI were retrospectively recruited from two independent centers (center1, training set, n = 115; validation set, n = 14; internal test set, n = 15; center 2, external test set, n = 21).
The medmanba was used to extract image features on the pre-contrast, arterial phase, portal venous phase, and hepatobiliary phase of MRI. These features combined with clinical factors (such as Neutrophils, lymphocytes, and serum tumor markers), and classified by a linear layer.
For comparison, a DL model was constructed by removing clinical factors, and a clinical model was established by the random forest selection on the clinical features.
Model performance was evaluated with the area under the receiver operating characteristic curve (AUC).
Gradient-weighted class activation mapping (Grad-CAM) heatmaps were used to show the focus area in predicting PNI.
Results or Findings: The PNI positive rate was 42.4% (61/144) in center 1 and 28.6% (6/21) in center 2.
On the internal test and external test set, the combined model showed the highest AUC of 0.944 and 0.844. The DL model achieved the moderate AUC of 0.926 and 0.833. The performance of clinical model is relatively low, with AUCs of 0.852 and 0.711.
Grad-CAM showed the DL model focused on the solid component of the tumor, especially the margin area.
Conclusion: MRI based DL model can accurately predict PNI-positive ICC, and the tumor margin area may have important indications for the model.
Limitations: Retrospective study; limited sample size.
Funding for this study: National Natural Science Foundation of China (82271958)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The Institutional Review Board of The First Affiliated Hospital, Sun Yat-sen University(approval number: [2023]014)

Author Block: X. Zhou, M. Chen, S-T. Feng; Guangzhou/CN
Purpose: To preoperatively predict isocitrate dehydrogenase 1 (IDH1) mutation and perineural invasion (PNI) of intrahepatic cholangiocarcinoma (ICC) based on the Gd-EOB-DTPA-enhanced MRI and MRI-derived habitat imaging to improve the reliability and interpretability.
Methods or Background: A total of 129 ICC patients with Gd-EOB-DTPA-enhanced MRI before resection between 2018 and 2024 were collected, and randomly assigned to training set and the test set in a ratio of 7:3. IDH1 mutation and PNI status were assessed on pathologic tissue slides. Clinical information and MRI features were qualitatively and quantitatively collected. Matchable tumors in the pre- and post-enhancement T1 mapping images were manually outlined for habitat analysis and divided into five habitats based on kmeans clustering (Habitat 1-5). A combined nomogram model was constructed based on clinical features, MRI features and habitat fraction. The diagnostic accuracy was evaluated using the area under the receiver operating characteristic curves (AUCs).
Results or Findings: The IDH1 nomogram model consists of age, T2 central brightness, liver ADC value, tumor T1 value reduction rate and percent of Habitat 4, with AUCs of 0.926 and 0.924 in the training and validation sets. The PNI nomogram model consists of CEA, tumor location, intrahepatic bile duct dilation and percent of Habitat 1, with AUCs of 0.854 and 0.896 in the training and validation sets. By mapping the habitats to multi-sequence MRI, Habitat1 is located predominantly at the edge of the tumor, with signals suggestive of a parenchymal component representing the aggressive edge of the tumour. Habitat 4 is located intratumorally, with signals suggestive of an intratumoral fibrotic area with little tumor component.
Conclusion: MRI and habitat imaging can noninvasively and preoperatively determining the IDH1 mutation and PNI of ICC with good accuracy and interpretability.
Limitations: Retrospective study, limited sample size.
Funding for this study: National Natural Science Foundation of China (82271958)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The Institutional Review Board of The First Affiliated Hospital, Sun Yat-sen University(approval number: [2023]014)

Author Block: K. Kadirli, A. Usta, S. Sahin, A. Cantürk, S. Özkan, H. Mutlu; Istanbul/TR
Purpose: Routine abdominal MRI often shows signal changes in the gallbladder on T1- and T2-weighted images, aside from gallstones. Differentiating between bile sludge and physiological bile concentration is key for accurate diagnosis, but literature shows overlapping signal characteristics with limited guidance. This study aims to identify specific MRI findings to distinguish between bile sludge and physiological bile concentration for improved diagnosis and clinical management.
Methods or Background: This retrospective study included patients with non-stone signal changes on upper abdominal MRI from January 2022 to April 2024, who had follow-up ultrasounds within one month. The study involved 42 patients in the sludge group and 44 in the non-sludge group according to ultrasound findings. Key parameters assessed were layering, fluid-fluid levels on T2, T1 signal characteristics, and signal loss on out-of-phase (OOP) sequences. Quantitative measures included the T2 abnormal signal/spleen ratio, ADC values, and the normal-bile/abnormal-signal ratios on T2 and ADC maps. Multivariate regression was performed on parameters with p < 0.05, and the model's diagnostic performance was evaluated.
Results or Findings: T1-weighted hypointensity and fluid-fluid levels were strongly associated with bile sludge (p < 0.001), while layering and OOP signal loss associated with normal bile(p<0,001). The best multivariate model, using T1W, fluid levels, and OOP signal loss, showed 83.72% specificity, 72.22% sensitivity, and 78.48% accuracy, with an AUC of 0.889.
Conclusion: An abnormal signal in the gallbladder lumen on MRI that does not exhibit OOP signal loss but shows fluid-fluid levels and T1-weighted hypointensity may be indicative of bile sludge.
Limitations: This study is limited by its retrospective design, which may introduce selection bias. Additionally, ultrasound was used as the gold standard, biochemical evaluation of the bile was not conducted.
Funding for this study: No funding
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee approval was obtained from the relevant institution.