Presentation Information

[POS-25]Prediction of graft loss in living donor liver transplantation in early postoperative period using machine learning

*takeru matsuura1, Shingo Iwami1 (1. Nagoya Univ. (Japan))

Keywords:

Liver transplantation,Machine learning

The liver plays an important role in the human body, performing a variety of vital functions. The loss of function of that liver means death for the human body, and liver transplantation is the only option for the survival and reintegration of such patients. However, this procedure carries the risk of graft loss. Therefore, prevention of postoperative graft loss is extremely important, but statistical indicators to predict graft loss have yet to be put into practical use in clinical practice.In this study, we used cohort data from 748 patients who underwent LDLT provided by Kyushu University Hospital and applied machine learning methods to predict graft loss. As a results, we developed a machine learning model to predict early graft loss (within 180 days postoperatively) with better performance than conventional models and to identify biomarkers highly correlated with graft loss. Using the model, we also stratified a highly heterogeneous patients into five groups. By focusing on survival time, we next categorized the patients into three groups: G1, G2, and G3+G4+G5. Notably, we identified G2 as a distinct population with a different survival time than the early graft loss group (G1) and the long-term survival group (G3+G4+G5). Additionally, by proposing a hierarchical prediction method, we developed an approach to distinguish these populations using data up to 30 days postoperatively.Our findings enabled us to establish a prediction method that effectively identifies patients at high risk of graft loss following liver transplantation. This research is expected to contribute to the development of a medical system that facilitates the early detection and prevention of graft loss, ultimately improving patient outcomes.