Prognose coronary heart diseases through sphygmogram analysis and SVM classifier

Abstract

A method of using statistical analysis on site-sampled sphygmogram data sets and support vector machines classifier to diagnose coronary heart disease is proposed. The hemodynamic parameters derived from sphygmogram reflect the status of human cardiovascular system. Based on homodynamic parameters, the dimension reduction methods and a modified support vector machines classifier are applied to meliorate prognosis sensitivity and specificity. The test results on clinical coronary heart disease patients show that this method has obvious advantages over existing classifier method in the captioned application.

Hierarchical Probabilistic Support Vector Machine for Detecting Cardiovascular Diseases

“…(8) Here, N + is the number of patients who have the d th CVD and N -is the number of patients who do not have the d th CVD in the total training dataset. It is emphasized that Level 1d and Level 2d PSVM are constructed in the same procedure with Level 3d PSVM, except only that the dimension of input feature space are diverse.…”

“…Concretely, decision tree can fast deduce conclusion from HDPs according to exclusively expressed medical knowledge [7], but calculations might get extremely complex while handling uncertain values; support vector machine shows high accuracy in detecting CVDs [8], but has disadvantage of time consuming and contradicting to the doctors' clinical diagnosis procedure since it is one-layer classifier. In practice, doctor normally ranks all parameters and selects specified ones with most pertinence as diagnostic basis in accordance with different diseases, which makes the reasoning procedure representing "hierarchically" character.…”