Real-time prediction of cardiovascular diseases using reservoir-computing and fusion with electronic medical record
Sudarsan Sadasivuni, Vasundhara Damodaran, Imon Banerjee, Arindam Sanyal · 2022 IEEE 4th International Conference on Artificial Intelligence Circuits and Systems (AICAS) · 2022
Cardiovascular diseases (CVDs) are a leading cause of death in USA and globally, but many people suffering from CVDs are asymptomatic in the early stages leading to reduced awareness, and less chances of managing the disease. This work presents a potential solution for at-home monitoring by leveraging predictive power of artificial intelligence (AI) for developing a fusion framework that combines patient electrocardiogram (ECG) and electronic medical record (EMR) for predicting risk of CVDs at an early stage. To improve energy-efficiency of wearable ECG sensor, in-sensor analog reservoir-computing is proposed that precludes need for front-end digitization and transmission of raw sensor data. The fusion framework predicts ischemic heart disease (I20–I25 ICD codes) with area under the receiver operating characteristic (AUROC) of 0.91, and other heart diseases (I30–I52 ICD codes) with AUROC of 0.95 which is better than state-of-the-art while not requiring laboratory test results.