MedTS-TTT: Test-Time Training for Medical Time Series Classification

2026-06-19

AI Summary

This work addresses the challenge of distribution shift in medical time series—such as EEG and ECG—caused by inter-subject variability, which severely degrades model generalization to unseen individuals. To tackle this issue, the authors propose MedTS-TTT, a novel framework that integrates a closed-loop self-aligned test-time training mechanism (CLSA-TTT) with a gated convolutional backbone (GCB). This design enables token-level self-supervised learning objectives, facilitating rapid single-step weight adaptation and dynamic local modeling during inference, thereby eliminating the computational overhead of inner-loop iterations in conventional test-time training (TTT) approaches. Evaluated across four public medical time-series datasets under twelve benchmarks, MedTS-TTT achieves top-1 performance in eleven tasks and consistently outperforms nine state-of-the-art baselines.

Abstract

Medical time series (MedTS) signals such as electroencephalography (EEG) and electrocardiography (ECG) support many clinical applications. However, substantial subject-level heterogeneity often induces subject-level distribution shift, causing a fixed parameter set to generalize poorly to unseen individuals. Compared with domain adaptation methods that often depend on extra adaptation components or target-batch statistics, Test-Time Training (TTT) provides a more practical solution for sequential clinical data by enabling online adaptation from unlabeled test samples. However, many representative TTT methods require iterative inner-loop optimization, increasing test-time overhead. In this paper, we propose MedTS-TTT, a test-time training framework for medical time series modeling. MedTS-TTT is built upon Closed-Loop Self-Alignment Test-Time Training (CLSA-TTT) and a Gated Convolutional Backbone (GCB). CLSA-TTT constructs a token-level self-supervised target and performs a single-step fast-weight update for intra-layer closed-loop alignment, enabling rapid sample-wise adaptation without iterative inner-loop optimization. GCB combines CLSA-TTT-based fast adaptation and token-level fusion with a gated convolutional branch to balance local dynamic modeling and information-flow control. On 4 public datasets (2 EEG and 2 ECG) with subject-independent splits, MedTS-TTT achieves 11 top-1 rankings out of 12 evaluations across 9 baselines and 3 metrics. The code is publicly available at https://github.com/mingzhi-c/MedTS-TTT.

Problem

• medical time series
• subject-level heterogeneity
• distribution shift
• generalization
• clinical data

Innovation

• Test-Time Training
• Medical Time Series
• Closed-Loop Self-Alignment
• Gated Convolution
• Distribution Shift