Turkish Journal of Electrical Engineering and Computer Sciences




Cardiac diseases (CDs) are one of the leading causes of the growing global mortality rate. Early detectionof CDs is necessary to avoid a high increase in the mortality rate. Machine learning-based computer-aided diagnosisof CDs using various physiological signals has recently been used by researchers. Since pulse plethysmograph (PuPG)signal contains a wealth of information about cardiac pathologies, therefore, this paper presents an expert system designfor the automatic diagnosis of cardiac disorders like hypertension, dilated cardiomyopathy and myocardial infarctionusing a novel fingertip PuPG signal analysis. The proposed system first performs signal denoising of raw PuPG sensordata using discrete wavelet transform (DWT). After signal segmentation, it extracts discriminant and simplest time-domain features, which are used to perform the detection of normal and abnormal subjects through a support vectormachine (SVM) classifier. The proposed detection and classification systems are tested using 10-fold cross-validationwhich yielded an average accuracy of 98.90%, sensitivity of 100.00%, and specificity of 98.02% for detection (normalvs. abnormal) experiments with only four features and an average accuracy of 97.57% for the multiclass problem usingfive computationally inexpensive features. Comparative analysis with existing methods based on electrocardiogram,photoplethysmograph, and phonocardiogram revealed that the proposed system has high e?iciency in terms of CDdetection with very low computational complexity. The findings of this work provide insights into the contributionof PuPG signal analysis towards accurate detection of cardiac disorders through innovative, low cost, and noninvasivemethods. Such a system with mobile cardiac health monitoring could be used as a counterpart or second opinion withclinical diagnoses and provide patients with additional but subtle indicators of varying heart dynamics.


Pulse plethysmograph (PuPG), biomedical signal analysis, support vector machine, feature extraction, discrete wavelet transform, cardiac disorders

First Page


Last Page