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

Area of Science:

• Biomedical Engineering
• Cardiovascular Health
• Neurology

Background:

• Cerebral infarction (CI), or stroke, results from reduced blood supply to the brain, necessitating early detection to prevent severe complications.
• Current diagnostic methods like CT, MRI, and ultrasound are effective but are time-consuming and costly.
• Photoplethysmography (PPG) signals from wearable devices show promise for non-invasive cardiovascular disease diagnosis.

Purpose of the Study:

• To develop and validate an automated, fiducial point-independent framework for predicting cerebral infarction using PPG signals.
• To assess the accuracy of the proposed method in classifying individuals with and without CI.
• To establish a simple, cost-effective approach for early CI detection and raise awareness for timely treatment.

Main Methods:

• Utilized a publicly available dataset comprising PPG and other physiological data from 219 individuals.
• Developed an automated framework to extract entropy features from higher-order PPG derivatives.
• Employed Coarse Gaussian Support Vector Machine (SVM) for classification of CI subjects.

Main Results:

• Achieved high accuracy in predicting cerebral infarction, with specific validation and test accuracies reported ($$ and $$).
• The fiducial point-independent approach demonstrated sufficient diagnostic capability.
• The method successfully classified CI subjects from normal individuals.

Conclusions:

• The proposed PPG-based framework provides a simple, automated, and inexpensive method for early cerebral infarction detection.
• This approach can aid in identifying individuals at risk, promoting awareness, and facilitating timely treatment to avoid major health disorders.
• Entropy features extracted from PPG derivatives are valuable for predicting CI pathology.