Development of a Radiomics-Based Model to Predict Graft Fibrosis in Liver Transplant Recipients: A Pilot Study

F. Qazi Arisar¹, E. Salinas-Miranda², H. Ale Ali², K. Lajkosz³, C. Chen¹, A. Azhie¹, G. M. Healy², D. Deniffel², M. A. Haider², M. Bhat¹

¹Ajmera Transplant Program, University Health Network, Toronto, ON, Canada, ²Lunenfeld Tanenbaum Research Institute, Toronto, ON, Canada, ³Department of Biostatistics, University Health Network, Toronto, ON, Canada

Meeting: 2022 American Transplant Congress

Abstract number: 323

Keywords: Area-under-curve (AUC), Graft failure, Prediction models, Radiologic assessment

Topic: Clinical Science » Liver » 62 - Liver: Large Data and Artificial Intelligence

Session Information

Session Name: MELD Allocation and Large Data

Session Type: Rapid Fire Oral Abstract

Date: Monday, June 6, 2022

Session Time: 5:30pm-7:00pm

Presentation Time: 6:50pm-7:00pm

Location: Hynes Room 313

*Purpose: Recurrent fibrosis complicates 40% of Liver transplants (LT), compromising long-term survival. We evaluated the ability of a model combining radiomic features on CT scans alongside longitudinal clinical variables within the first 6 months post-LT to flag patients at risk of developing significant graft fibrosis in long term. We hypothesized that radiomic features (subtle perfusion, biliary and parenchymal changes) early on post-LT could provide insight into the long-term life span of the graft, beyond the longitudinal clinical information.

*Methods: Computed Tomography of 254 patients at 3-6 months post-LT between 2009 – 2018 were analyzed. Volumetric radiomic features were extracted from portal phase using PyRadiomics, an Artificial Intelligence-based tool. The primary endpoint was advanced graft fibrosis (≥F2 on transient elastography or histopathology). A 5-fold cross-validated LASSO model using clinical and radiomic features was developed.

*Results: 75 patients (29.5%) developed ≥F2 fibrosis by a median of 19 (4.3 – 121.8) months from transplant. The original first order maximum calculated at venous phase, a radiomic feature reflecting venous perfusion, significantly predicted future graft fibrosis (OR 0.52, 95%CI 0.38 – 0.71, p<0.001). Among the clinical variables, primary etiology of alcohol (OR 4.86, 95% CI 1.43 - 17.48, p=0.012) donor age (OR 1.04, 95%CI 1.01 - 1.07, p=0.003), recipient age at transplant (OR 0.94, 95%CI 0.9 - 0.98, p=0.003), recurrence of primary etiology (OR 5.28, 95%CI 2.11 - 14.2, p=0.001), brain-dead donor (OR 0.16, 95%CI 005 - 0.48, p=0.001), tacrolimus use at 3 months post-LT (OR 0.26, 95%CI 0.1 - 0.64, p=0.004) and APRI score at 3 months post-LT (OR 2.07, 95%CI 1.39 - 3.26, p=0.001) were significantly associated with ≥F2 fibrosis on multivariate analysis. Our model combining clinical and radiomic features predicted graft fibrosis with AUC of 0.814 (95%CI 0.749 - 0.90), sensitivity of 0.863 and specificity of 0.559.

*Conclusions: Our pilot study provides proof-of-principle that a combination of radiomic, clinical and laboratory features early post-transplant can prognosticate advanced graft fibrosis. This tool would serve to individualize the management of liver transplant recipients, by addressing any modifiable risk factors.

To cite this abstract in AMA style:

Arisar FQazi, Salinas-Miranda E, Ali HAle, Lajkosz K, Chen C, Azhie A, Healy GM, Deniffel D, Haider MA, Bhat M. Development of a Radiomics-Based Model to Predict Graft Fibrosis in Liver Transplant Recipients: A Pilot Study [abstract]. Am J Transplant. 2022; 22 (suppl 3). https://atcmeetingabstracts.com/abstract/development-of-a-radiomics-based-model-to-predict-graft-fibrosis-in-liver-transplant-recipients-a-pilot-study/. Accessed May 8, 2025.

« Back to 2022 American Transplant Congress