Multimodal Deep Fusion Model and Assistant Diagnostic Software for ECG Arrhythmia Classification

Cardiovascular diseases, particular arrhythmias, are significant health concerns worldwide. Accurate and prompt diagnosis of arrhythmias is crucial for patient care. Traditional methods of arrhythmia detection heavily depend on the expertise of physicians, which can be laborious. Our research introduces a multimodal deep fusion model that amalgamates diverse data streams-raw ECG segments, statistical and frequency features-within an attention based convolutional neural network architecture. Utilizing the MIT -BIH Arrhythmia Dataset, the model was trained and evaluated. According to the results, the proposed approach is able to make prediction with the accuracy of 98.66% on arrhythmia classification. Additionally, an assistant diagnostic software integrating the proposed model, is developed to provide a user-friendly interface to facilitate swift and reliable diagnostic insights.