Dual PPG-Based Blood Flow Abnormality Detection Using Deep Learning

This paper presents a novel system for predicting nonuniform blood flow using dual Photoplethysmography (PPG) signals and deep learning. The hardware comprises two PPG sensors strategically placed on the finger and thumb muscle, connected to an ESP32 node MCU board for data acquisition. The captured signals are transmitted via Wi-Fi to a computer for analysis. A Convolutional Neural Network (CNN) classifier processes the dual PPG data to categorize blood flow as normal or abnormal, providing a confidence level for each prediction. The model, trained on a dataset of 960 samples from 22 individuals, achieved an accuracy of 87.3% in distinguishing between normal and abnormal blood flow patterns. Additional performance metrics, including precision (84.8%), recall (90.3%), and F1-score (87.5%), demonstrate the system's effectiveness. This approach offers a promising non-invasive method for early detection of blood flow abnormalities, potentially improving cardiovascular health monitoring.