DeepArrhythmia: AI Framework for ECG Beat Classification

DeepArrhythmia is a multimodal framework designed for classifying ECG arrhythmias at the beat level, based on tool integration. It merges raw ECG data with waveform images, identifies R peaks, and generates organized predictions at the beat level. This framework separates physiological measurements from the integration of evidence by utilizing dedicated tools for localizing beats, extracting numerical rhythm morphology, and performing morphology-centered textual analysis. By leveraging segment-level confidence, it effectively navigates between tools, overcoming the challenge of analyzing beats as standalone occurrences by integrating the context of multiple beats.