Mortality Prediction Based on Echocardiographic Data and Machine Learning: CHF, CHD, Aneurism, ACS Cases

This paper represents the research results of echocardiographic study for early prediction of mortality. The classification task is solved by analyzing the echocardiographic data from medical information system. Echocardiographic data of 90000 hearts condition extracted directly from medical information system were analyzed. The considered echocardiographic studies were conducted for patients with CHF, CHD, hypertension, heart arrhythmia, valvular heart disease, autoimmune disease, congenital heart defect, cardiomyopathy, endocrine disease, heart failure, mixed connective tissue disease, cancer, aneurism. Using machine learning methods and neural networks it is possible to make an early prediction of mortality based on instrumental echocardiographic study. It can be offered to clinicians as support for accurate, reasonable saving clinical decisions with minimization risks for patient's health. The classification task mortality prediction is solved by machine learning methods with 97% ROC curve. The simple echocardiographic test results like FV Simpson, systolic volume, the valves condition, the condition of the ascending aorta and other echocardiographic data are used as predictors. Such a simple approach to solving critical tasks can make the method widely used in clinical practice.