Metaheuristic-based dimensionality reduction combined with machine learning classifiers can achieve high accuracy in classifying cardiovascular disease from PPG signals.
One of the most useful methodologies which provide a specific waveform showing the pulsating peripheral blood flow in a non-invasive manner is Photoplethysmography (PPG). The design, application and implementation of a PPG system are quite inexpensive and have a very easy maintenance. Without having direct contact with the surface of the skin, PPG can easily take the measurements. Therefore, PPG has a good medical competency and due to its widespread availability, it has a lot of advantages. A PPG signal can sometimes be substituted or complemented by an Electrocardiography (ECG) signal as it can provide Heart Rate Variability (HRV) analysis. In this work, an in-depth analysis of classification of Cardiovascular Disease (CVD) is done with the help of Capnobase dataset. Initially, metaheuristic optimization algorithms are utilized as dimensionality reduction techniques and then the dimensionally reduced values are classified with the help of different classifiers for the classification of CVD. The results show that for the PPG normal cases, a high classification accuracy of 99.48% is obtained when Chi square Probability Density Function (PDF) optimized values are classified with Artificial Neural Networks (ANN) and a second highest classification accuracy of 98.96% is obtained when Chicken swarm optimized values are classified with Naïve Bayesian Classifier (NBC). Similarly when the PPG abnormal cases or PPG with CVD cases are concerned, a high classification accuracy of 99.48% is obtained when Chi square PDF optimized values are classified with Logistic Regression and a second highest classification accuracy of 98.96% is obtained when Chi square CDF optimized values are classified with Gaussian Support Vector Machine (SVM) and when Chicken swarm optimized values are classified with NBC.
Prabhakar et al. (Tue,) studied this question.