The Cardio-AI system, utilizing a 1D-CNN deep learning algorithm, achieved 98.13% overall accuracy for real-time ECG arrhythmia classification across 20,261 independent test heartbeats.
Does the Cardio-AI system accurately detect and classify cardiac arrhythmias from ECG signals?
The Cardio-AI system demonstrates high accuracy (98.13%) in real-time ECG arrhythmia detection and classification using a 1D-CNN model and custom hardware.
Cardiac arrhythmias represent a critical healthcare challenge, affecting millions worldwide and requiring rapid, accurate diagnosis for effective clinical intervention. This comprehensive thesis presents Cardio-AI, an integrated system combining custom hardware signal acquisition with advanced deep learning algorithms for real-time electrocardiogram (ECG) arrhythmia detection and classification. Our approach implements a custom Arduino-based ECG acquisition module utilizing an AD8232 sensor with dual transmission pathways (USB and Bluetooth). The system was optimized to acquire signals at a 250 Hz sampling rate, ensuring real-time stability. On the software side, a 1D-Convolutional Neural Network (1D-CNN) was developed and trained on the MIT-BIH Arrhythmia Database. The system classifies heartbeats into seven clinically relevant categories: Normal, Left Bundle Branch Block (LBBB), Right Bundle Branch Block (RBBB), Atrial Premature Contraction (APC), Premature Ventricular Contraction (PVC), Supraventricular Ectopic Beat (SVE), and Fusion beats. The 1D-CNN model achieves 98.13% overall accuracy on a test set of 20,261 independent heartbeats, with an F1-score of 0.95 for critical PVC arrhythmias. A novel feature of this implementation is the severity-hierarchy algorithm that prioritizes the most dangerous arrhythmia detected within an ECG strip, significantly enhancing clinical relevance. The complete system—comprising custom hardware, the AI model, and a user-friendly desktop application—represents a cost-effective, deployable solution for cardiac rhythm monitoring.
Khaled Hussam Burhan (2026) studied Cardiac arrhythmias. Cardio-AI (1D-CNN deep learning algorithm and custom ECG hardware) was evaluated on Arrhythmia detection and classification accuracy. The Cardio-AI system, utilizing a 1D-CNN deep learning algorithm, achieved 98.13% overall accuracy for real-time ECG arrhythmia classification across 20,261 independent test heartbeats.