Prediction Model of Left Ventricular Function after Transplantation Based on Multivariate Linear Regression

Evaluating donor heart function during ex situ heart perfusion (ESHP) is crucial for determining transplantation suitability. Cardiac Index (CI) is a key hemodynamic parameter widely used to assess left ventricular function. In this study, a predictive model for CI was developed using ESHP-derived physiological parameters. To develop an efficient predictive model for CI, a feature selection strategy was employed, incorporating variance filtering, maximal information coefficient (MIC) analysis, and Affinity Propagation clustering. A multiple linear regression model was initially constructed using the selected features, followed by optimization through best subsets regression to enhance model simplicity and clinical interpretability. The final model, based on a reduced set of critical features, demonstrated strong predictive capability while maintaining practical applicability. The proposed model effectively predicts posttransplant cardiac function using ESHP-derived parameters, offering a clinically actionable decision tool for heart transplantation. Future studies will focus on improving predictive accuracy and validating the model in clinical settings.