EXPLAINABLE AI BASED ARRHYTMIA MONITORING SYSTEM

Abstract: Cardiovascular diseases are a major global health concern, creating a need for intelligent and realtime cardiac monitoring systems. Conventional electrocardiogram (ECG) analysis often depends on specialists and may not provide immediate support in emergency or remote settings. This paper presents CardioSense AI, an explainable real-time ECG monitoring platform that combines embedded sensing, deep learning, automated reporting, and a hospital-grade dashboard for rapid rhythm assessment. The system uses an AD8232 sensor with an ESP32 microcontroller for live ECG acquisition. A one-dimensional convolutional neural network (1D-CNN) directly classifies ECG signals into Normal Rhythm, Bradycardia, Tachycardia, and PVC-type Arrhythmia. To improve transparency, Grad-CAM based Explainable AI (XAI) highlights waveform regions influencing the model prediction. The platform also provides live ECG streaming, alerts, confidence indicators, and simulation support using the MIT-BIH Arrhythmia Database. In addition, an automated report generation module enables downloadable clinical summaries after each session. Experimental results show that the proposed system achieves effective real-time classification with interpretability and practical usability. CardioSense AI offers a scalable solution for telecardiology, bedside monitoring, and preventive healthcare applications.