Classification Model of Heart Transplant Outcomes Based on Features of Left Ventricular Functional Geometry

Abstract

The function of the transplanted heart can be significantly affected by acute allograft rejection, chronic rejection, high blood pressure. These factors may induce cardiac remodelling with further adverse consequences for the patients. Dynamic change in the configuration of the left ventricle (LV) from end diastole to end systole (LV functional geometry) is an important factor of the heart pump function. The objective of this study is to evaluate the time dependent changes in parameters of LV functional geometry in the transplanted heart and to assess relations between the changes and adverse outcomes of the heart transplantation (HT). We used linear discriminant analysis (LDA) to build classification models based on either the standard echocardiographic parameters of LV systolic function and global longitudinal strain (GLS) or LV function geometry indexes. The training set for model building included data from patients with different degrees of systolic dysfunction. Using the models, we retrospectively classified data from 31 patients after orthotropic HT. In contrast to the LDA models based on the standard echocardiographic characteristics and GLS, the model based on the LV functional geometry data showed high accuracy in predicting allograft rejection and development of the heart failure in the HT patients. © 2018 Creative Commons Attribution.