MambaCapsule: Towards Transparent Cardiac Disease Diagnosis with Electrocardiography Using Mamba Capsule Network

TL;DR

MambaCapsule addresses the critical need for transparent ECG-based arrhythmia diagnosis by integrating a multi-scale Mamba encoder with a Capsule network and a signal reconstruction module. The approach provides not only high classification accuracy on MIT-BIH and PTB datasets, but also interpretable explanations through reconstructed signals and capsule-level reasoning. Key contributions include a fusion states SSM backbone for multi-scale feature extraction, a capsule-based decoder with a reconstruction branch, and explainability analyses such as feature relation mining and disease manifestation inspection. The work demonstrates the feasibility of accurate, explainable ECG classification with potential clinical impact, while acknowledging training-time and memory demands that motivate future optimization.

Abstract

Cardiac arrhythmia, a condition characterized by irregular heartbeats, often serves as an early indication of various heart ailments. With the advent of deep learning, numerous innovative models have been introduced for diagnosing arrhythmias using Electrocardiogram (ECG) signals. However, recent studies solely focus on the performance of models, neglecting the interpretation of their results. This leads to a considerable lack of transparency, posing a significant risk in the actual diagnostic process. To solve this problem, this paper introduces MambaCapsule, a deep neural networks for ECG arrhythmias classification, which increases the explainability of the model while enhancing the accuracy.Our model utilizes Mamba for feature extraction and Capsule networks for prediction, providing not only a confidence score but also signal features. Akin to the processing mechanism of human brain, the model learns signal features and their relationship between them by reconstructing ECG signals in the predicted selection. The model evaluation was conducted on MIT-BIH and PTB dataset, following the AAMI standard. MambaCapsule has achieved a total accuracy of 99.54% and 99.59% on the test sets respectively. These results demonstrate the promising performance of under the standard test protocol.

Figures (7)

• Figure 1: The architecture of MambaCapsule. The model consists of three main parts:the feature network, the capsule network and the reconstruction network. We propose a fusion states SSM to construct backbone of the feature network.
• Figure 2: Training set about learning rate and $m^{+}$
• Figure 3: The F1-score of different states
• Figure 4: The F1-score of different layers
• Figure 5: The F1-score with different encoders.