[2605.05214] MedMamba: Recasting Mamba for Medical Time Series Classification

Electrical Engineering and Systems Science > Signal Processing arXiv:2605.05214 (eess) [Submitted on 17 Apr 2026] Title:MedMamba: Recasting Mamba for Medical Time Series Classification Authors:ZhengXiao He, Huayu Li, Xiwen Chen, Janet M Roveda, Jinghao Wen, Siyuan Tian, Ao Li View a PDF of the paper titled MedMamba: Recasting Mamba for Medical Time Series Classification, by ZhengXiao He and 5 other authors View PDF HTML (experimental) Abstract:Medical time series, such as electrocardiograms (ECG) and electroencephalograms (EEG), exhibit complex temporal dynamics and structured cross-channel dependencies, posing fundamental challenges for automated analysis. Conventional convolutional and recurrent models struggle to capture long-range dependencies, while Transformer-based approaches incur quadratic complexity and often introduce redundant interactions that are misaligned with the intrinsic structure of physiological signals. To address these limitations, we propose MedMamba, a principle-driven multi-scale bidirectional state space architecture tailored for medical time series classification. Our design is guided by three key inductive biases of physiological signals: spatial centralization, multi-timescale temporal composition, and non-causal contextual dependency. These principles are instantiated through a lightweight channel-mixing module for cross-channel reparameterization, multi-scale convolutional tokenization for temporal decomposition, and bidirectional Mamba bloc...