Related Experiment Video
Updated: Jun 3, 2025
08:49
Author Spotlight: Enhancing Graft Viability Assessment Through Quantitative Metrics and Innovative Reservoir Systems
Published on: August 2, 2024
661
Machine Learning Algorithms in Controlled Donation After Circulatory Death Under Normothermic Regional Perfusion: A
Rafael Calleja1, Marcos Rivera2, David Guijo-Rubio2
1Hepatobiliary Surgery and Liver Transplantation Unit, Maimonides Biomedical Research Institute of Cordoba (IMIBIC), Hospital Universitario Reina Sofía, University of Córdoba, Córdoba, Spain.
Transplantation
|January 9, 2025
Summary
Machine learning models accurately predict graft survival in liver transplants from controlled donation after circulatory death (cDCD) using normothermic regional perfusion (NRP). A new risk score improves donor-recipient matching for cDCD-NRP liver grafts.
Area of Science:
- Hepatology and Transplant Surgery
- Medical Informatics and Machine Learning
Background:
- Existing risk scores for graft loss in controlled donation after circulatory death (cDCD) show limitations in the Spanish population.
- Most cDCD livers in Spain are recovered using normothermic regional perfusion (NRP), necessitating region-specific predictive models.
- The need for improved risk stratification to optimize donor-recipient matching for cDCD-NRP liver transplants is critical.
Purpose of the Study:
- To evaluate machine learning classifiers for predicting graft survival in cDCD livers recovered with NRP.
- To develop a novel risk stratification score integrated into a donor-recipient matching system.
Main Methods:
- Retrospective multicenter cohort study involving 539 donor-recipient pairs of cDCD livers procured via NRP.
- Evaluation of multiple machine learning models: logistic regression, ridge classifier, support vector classifier, multilayer perceptron, and random forest.
- Analysis of 20 donor, recipient, and NRP variables to predict 3- and 12-month graft survival rates.
Main Results:
- Logistic regression demonstrated the highest predictive performance for graft survival at both 3 months (AUC=0.82) and 12 months (AUC=0.83).
- A donor-recipient matching system was proposed, incorporating the Model of End-Stage Liver Disease (MELD) score and the newly developed cDCD-NRP risk score.
- The proposed risk score showed satisfactory performance within the study cohort.
Conclusions:
- The developed risk score shows significant potential for supporting liver allocation decisions in cDCD-NRP grafts.
- The machine learning methodology offers a promising approach for risk stratification in liver transplantation.
- External validation is recommended, and this approach may be applicable to other geographical regions.
