Does the CardioNet deep learning framework accurately predict CVD risk factors using PPG data in subjects with known risk factors?
A lightweight deep learning framework using PPG data demonstrated high accuracy (97%) for detecting common cardiovascular risk factors, offering a potential non-invasive diagnostic tool.
The recent progressions in semiconductor and computing technology have empowered the PPG utilization in medical diagnosis. This paper presents a reconfigurable deep learning framework `CardioNet' for early diagnosis of cardiovascular risk factors or most common diseases (such as diabetes, hypertension, cerebrovascular, cerebra-infraction) using the PPG data. The proposed model has a light-weight architecture, designed by exploiting the deep learning framework of convolutional neural network, exhibiting inherent capability of feature extraction, thereby, eliminating the cost effective steps of feature selection and extraction. The performance demonstration of the proposed model is done on a healthy dataset comprising 657 data segments of 219 subjects holding records of common CVD risk factors (diabetes, hypertension, cerebrovascular, cerebra-infraction). The obtained results of an overall accuracy of 97% for diagnosis of CVD risk factors, show the efficiency of the proposed model for real-time usability. The clinical significance of this work to provide an accurate and non-invasive method for early diagnosis and monitoring of cardio-risk factors.
Panwar et al. (Tue,) studied this question.