Deriving ECG Features from PPG Signals Using Machine Learning Approaches

Electrocardiogram (ECG) plays a crucial role in diagnosing cardiovascular diseases, while Photoplethysmography (PPG) offers a non-invasive alternative for continuous heart monitoring, especially in wearable technology. Leveraging PPG for ECG feature estimation can enhance early detection and proactive management of heart-related conditions. This study presents a cost-effective approach to estimating key ECG parameters by utilizing the time-domain characteristics of PPG. Statistical analysis was conducted to identify the most relevant PPG features for ECG estimation, which were then used to develop a predictive model using artificial neural networks, linear regression, and random forest algorithms. Experimental results on a benchmark dataset confirm that the proposed methodology achieves high accuracy in ECG parameter estimation. The proposed cost-effective device (approx. one thousand), integrating feature selection, ensures high classification accuracy, real-time monitoring, and timely alerts, enhancing practical usability.