CIsense: an automated framework for early screening of cerebral infarction using PPG sensor data

A cerebral infarction (CI), often known as a stroke, is a cognitive impairment in which a group of brain cells perishes from a lack of blood supply. The early prediction and evaluation of this problem are essential to avoid atrial fibrillation, heart valve disease, and other cardiac disorders. Different clinical strategies like Computerized tomography (CT) scans, Magnetic resonance imaging (MRI), and Carotid (ka-ROT-id) ultrasound are available to diagnose this problem. However, these methods are time-consuming and expensive. Wearable devices based on photoplethysmography (PPG) are gaining prevalence in diagnosing various cardiovascular diseases. This work uses the PPG signal to classify the CI subjects from the normal. We propose an automated framework and fiducial point-independent approach to predict CI with sufficient accuracy. The experiment is performed with a publicly available database having PPG and other physiological data of 219 individuals. The best validation and test accuracy of $$ and $$ are obtained after diagnosis with Coarse Gaussian SVM. The proposed work aims to extract cerebral infarction pathology by extracting relevant entropy features from higher order PPG derivatives for the prediction of CI and offers a simple, automated and inexpensive approach for early detection of CI and promotes awareness for the subjects to undergo further treatment to avoid major disorders.