AI-based Predictive Health Analytics Model for Early Diagnosis of Cardiovascular Disease using Wearable Sensor Data

Abstract

Cardiovascular disease (CVD) remains the leading cause of death worldwide, and early risk identification outside the clinic remains difficult because conventional diagnostic pathways rely on episodic clinical visits. Wearable devices equipped with photoplethysmography (PPG), single-lead electrocardiography (ECG), and accelerometry now make continuous, low-cost physiological monitoring possible, creating an opportunity for artificial intelligence (AI) to convert these passively collected signals into actionable cardiovascular risk assessments. This paper proposes a predictive health analytics framework that fuses heart-rate-variability (HRV), photoplethysmographic, and activity features extracted from wearable sensor streams into a continuous cardiovascular risk estimate. The framework applies two simple algorithms formulated with elementary mathematics: (1) a z-score-based HRV risk-scoring procedure that normalizes HRV features against an individual or population baseline and combines them into a composite risk score, and (2) a weighted multi-sensor fusion algorithm that combines ECG-, PPG-, and activity-derived risk scores using validation-derived weights and classifies the result into Low, Medium, or High cardiovascular risk using a decision threshold. We describe the proposed methodology and its mathematical formulation in detail and report illustrative example results across three summary tables and two graphs to demonstrate the expected reporting format for classification accuracy and the sensitivity-specificity trade-off across decision thresholds. The framework is designed to be lightweight enough for on-device or edge deployment on consumer wearables.