Quality-Aware Approach to Arrhythmia Detection Using CNN and Hybrid RNN Models

Recent arrhythmia detection methods focus on identifying irregular heart rhythms but often ignore ECG signal quality, which significantly affects accuracy. Factors like noise, artifacts, and baseline wander can lead to false diagnoses. Instead of filtering or improving signal quality, the presented methodology selects high-quality ECG signals as input for a hybrid CNN-RNN model to enhance accuracy. The proposed CNN-RNN algorithm uses a multi-class classification model to detect different arrhythmia types. CNNs extract spatial features from ECG images, while LSTM-GRU RNNs capture temporal patterns for better classification. The model, tested on datasets like MIT-Arrhythmia, supports grayscale, RGB, and spectrogram ECG formats. High-Quality ECG inputs improved accuracy to 98.02%, while low-quality signals dropped it to 51.45%. By ensuring only good-quality signals are used, the method significantly reduces false positives and negatives, making arrhythmia detection more reliable and accurate in real-world medical applications.