Heart Guardian: Tinyml Based Ventricular Arrhythmia Detection for Life Saving Treatment

Abstract

Cardiovascular diseases, notably ventricular arrhythmias, continue to pose a significant threat to global health. Timely detection and intervention are crucial for improving patient outcomes, prompting the development of HeartGuardian—a pioneering solution that leverages TinyML (Tiny Machine Learning) for life-saving ventricular arrhythmia detection through unobtrusive wearable devices. This project introduces a methodological breakthrough by designing and implementing a highly optimized TinyML model tailored for wearable devices. The model analyses real-time electrocardiogram (ECG) signals directly on the device, achieving a delicate balance between accuracy and resource efficiency. This innovation enables continuous monitoring without imposing substantial power or computational demands, making it feasible for widespread use. Heart Guardian’s primary focus is on providing proactive and continuous monitoring to enable early detection of ventricular arrhythmias. The embedded TinyML model ensures immediate analysis of ECG data, allowing for prompt alerts to healthcare professionals or caregivers upon detection of irregularities. This real-time response mechanism is critical for timely medical intervention and potentially life-saving measures. The ultimate aim of our project is to maintain data privacy and safeguard sensitive information of patients Striving for compliance with privacy regulations, HeartGuardian aims to build trust in the utilization of wearable health technology. Thus, the project provides an effective method for ventricular arrhythmia detection using the Arrhythmia Classification Dataset. The dataset is pre-processed by filtering signals, and models, including Logistic Regression and Efficient CNN, are trained. Validation and evaluation are conducted, ultimately leading to the prediction of arrhythmia.