Classification of Cardiac Abnormality using PPG Fiducial Parameters and Probabilistic Neural Network

In the current era among all the untimed or premature root of mortality throughout the world is cardiovascular disease or abnormalities (CVD). For an effective prevention to save the human life is its early-stage detection is much more important. The medical professionals consider the predictor of these cardiac abnormalities by considering the physiological biomedical signals, not only Electrocardiogram (ECG) but also Phonocardiogram (PCG) and photoplethysmogram (PPG). To monitor these anomalies typically ECG is used to visualize the electrical activity of the heart, which is quite important but tedious for continuous and too lengthy time monitoring. In contrast, optical PPG has gained the popularity because of its less-cost, wireless and very tiny handy size. The cardiac abnormalities such as Hypertension, Heart failure, Ischemia, Myocardial infarction, Chronic obstructive pulmonary disease (COPD) and many more is reflected in a PPG waveform as they disrupt volumetric blood flow rate in the body. Here our work proposes an automatic method for recognizing these abnormalities in PPG signal only without the need for ECG. The proposed method suggests that a probabilistic neural network is trained with fiducial parameters of large number of PPG samples with its respective classes as cardiac abnormalities. The trained PNN is then utilized to predict the class of unknown PPG samples for quick diagnosis. The accuracy of prediction is tested as more than 95%.

Automated summary

Cardiovascular disease or abnormalities are the leading cause of premature death worldwide. Early detection is crucial for effective prevention and saving lives. This study proposes an automatic method for recognizing cardiac abnormalities using only PPG signals, achieving an accuracy rate of over 95%.