Interpretable machine learning model for early complication prediction after split liver transplantation
Di Wang1, Jun-Yan Zhang1, Yan Xie1
1Department of Liver Transplantation, First Central Hospital of Tianjin Medical University, Tianjin 300380, China.
World Journal of Gastroenterology
|January 2, 2026
Summary
Split liver transplantation (SLT) can increase early complications, but machine learning identified key risk factors. Partial lobectomy of segment IV (IV PL) may reduce complications and improve recovery in SLT recipients.
Area of Science:
- Hepatobiliary surgery
- Transplant surgery
- Machine learning in medicine
Background:
- Split liver transplantation (SLT) expands donor availability but increases early postoperative complications (EPC) risk.
- Altered graft hemodynamics and extensive surgical transection contribute to EPC in SLT.
Purpose of the Study:
- Develop an interpretable machine learning framework to identify EPC risk factors in adult SLT recipients.
- Specifically focus on right tri-segment SLT to improve patient outcomes.
Main Methods:
- Retrospective analysis of 109 adult SLT recipients (37 with EPC).
- Utilized random forest, SVM, XGBoost, and logistic regression with SHapley Additive exPlanations for variable importance.
- Validated predictors via multivariate logistic regression and constructed a diagnostic nomogram.
Main Results:
- EPC occurred in 33.9% of recipients; random forest showed best predictive performance.
- Key predictors identified: log-transformed systemic immune-inflammation index (LnSII), albumin-to-fibrinogen ratio, MELD score, partial lobectomy of segment IV (IV PL), blood loss, and operation time.
- Nomogram achieved an AUC of 0.788; LnSII, MELD, IV PL, and blood loss were independent predictors.
Conclusions:
- Partial lobectomy of segment IV (IV PL) in right tri-segment SLT may reduce EPC and improve liver function recovery.
- LnSII, MELD score, IV PL, and blood loss provide a basis for individualized perioperative risk stratification in SLT.
