Feature Extraction and Use of Artificial Neural Networks for Classification of PPG Signals

Recent studies associated to IoTs and health sector show that equine heart rates can be obtained from the ventral midline using a non-invasive technique called photoplethysmography (PPG). It is measured with infrared light-emitting diodes and photodetectors and is associated with the pulsatile volume of blood in tissue. PPG is becoming increasingly commonly utilised as a result of developments in optical technology. The physical properties of blood vessel walls and artery pulsatile function are impacted by ageing and cardiovascular events. Techniques assess the mechanical characteristics of arteries, but their broad clinical use and utility as risk stratification tools are limited by their theoretical, technical, and practical shortcomings. This study tries to evaluate the possibility of classification of PPG signals using Artificial Neural Networks (ANNs) into different clinical classes or states and the best possible combination of features that can be used for such classification. During the study PPG signals from 35 subjects or individual, each with 50–60 signals and 300 samples (6 seconds of recording) at a 50 sample/second sampling rate, make up the dataset for analysis. During the study we observed the use of different signal features like mean, root mean, mean peak height, mean time duration between peak etc. helped to classify the PPG signals into different clinical states. The results of the classification using ANNs demonstrated that ANNs were useful in differentiating between five states. Every ANN model evaluated during the training phase demonstrated significant validation accuracy (> 98 %), and also every ANN model tested during the testing phase demonstrated significant testing accuracy (> 97 %). The results of the study indicate that the use of the given features included in the study along with the ANNs models can have a huge potential in effective classification and detection of possible clinical states of heart functions or cardiovascular functions, which finally can help in early detection and medication of possible anomaly in ideal functions.