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

Tatiana Chumarnaya1, Sergey P. Mikhaylov2, Eduard M. Idov3, Olga Solovyova4
1Institute of Immunology and Physiology Ural branch of Russian Academy of Sciences, 2Urals State Medical University, 3Yekaterinburg Regional Clinical Hospital 1, 4Ural Federal University


Abstract

Cardiac transplantation is accepted as the best therapeutic option for end-stage heart failure patients who remain symptomatic despite optimal medical therapy. The function of the transplanted heart will be affected by acute allograft rejection, chronic rejection, high blood pressure and so on. All these factors may induce the remodeling of the left ventricle (LV) that will significantly affect the prognosis of heart transplantation (HT). The LV configuration changes from end diastole to end systole (LV functional geometry) are an important factor of the heart pump function optimization. The objective of this study was the assessment of HT outcomes using parameters of LV functional geometry. We used linear discriminant analysis (LDA) to build classification model which was based on classic echocardiographic parameters, LV function geometry indexes (such as: sphericity ; Fourier shape-power ; apical conicity, spatio-temporal heterogeneity) and speckle-tracking-derived LV longitudinal strain and strain rate. The training set consisted of three groups of patients with different degrees of LV systolic dysfunction: 1) 24 healthy volunteers with normal systolic function(CG); 2) 52 patients suffering from ischemic heart disease (IHDG) with preserved LV systolic function; 3) 35 patients with dilatational cardiomyopathy (DCMG) with significant LV systolic dysfunction. We examined 31 patients after orthotopic HT who had postoperative period from 1 to 9 years. All HT patients underwent pairs of endomyocardial biopsy (EMB) and echocardiographic evaluation on the same day. Group of HT patients were stratificated of our LDA model. One quarter of HT group were defined in CG. 44% and 31% HT patients were defined in IHDG and DCMG respectively. In 97% of HT patients the predictions of the LDA model coincided with the state of the allograft obtained by EMB. Our LDA model with parameters of LV functional geometry can facilitate the diagnosis and reveal deterioration of the allograft condition.