| Abstract: |
Cardiovascular disease (CVD) remains the leading cause of mortality worldwide, and delayed diagnosis continues to undermine early intervention efforts. This empirical study evaluates the diagnostic accuracy and real-world clinical integration of an artificial intelligence (AI)-enabled home monitoring system designed to detect early indicators of cardiovascular disease using wearable electrocardiogram (ECG) and photoplethysmography (PPG) sensors. A prospective observational design was employed with 500 participants across three healthcare centers over a six-month period, comparing AI-generated risk classifications against clinical ECG diagnoses rendered by cardiologists. Data were analyzed using sensitivity, specificity, predictive values, F1-score, and area under the receiver operating characteristic curve (AUC). Results indicate that the AI home monitoring system achieved a sensitivity of 91.2%, specificity of 88.4%, and an AUC of 0.931, approaching the diagnostic performance of clinical-grade ECG (sensitivity 94.8%, AUC 0.960). Patient adherence declined from 88% in month one to 71% by month six, while time-to-diagnosis improved substantially, from 6.2 days to 3.9 days. Comparative analysis against five prior studies confirms an upward trajectory in AI-ECG diagnostic performance. These findings establish that AI-enabled home monitoring provides a clinically meaningful, scalable complement to in-clinic diagnostics, materially reducing diagnostic latency while revealing adherence-related barriers that must be addressed for sustained real-world adoption and long-term cardiovascular risk reduction. |
