Related Experiment Video
Updated: Jun 10, 2026
11:06
Bilateral Assessment of the Corticospinal Pathways of the Ankle Muscles Using Navigated Transcranial Magnetic Stimulation
Published on: February 19, 2019
MyoFormer: Anatomy-guided transformer for muscle-level pathological gait assessment
Zheng Fang1, Jia Li2, Xiaoheng Zhao1
1Faculty of Information, Liaoning University, No. 66 Chongshan Middle Road, Shenyang, China.
Medical & Biological Engineering & Computing
|June 8, 2026
Summary
MyoFormer, a novel transformer model, accurately identifies pathological gait by analyzing muscle-level movement. This technology aids in early medical diagnosis and rehabilitation by improving gait analysis.
Area of Science:
- Biomechanics
- Medical Technology
- Artificial Intelligence
Background:
- Pathological gait is a key early indicator for various medical disorders.
- Accurate gait analysis is crucial for diagnosis and rehabilitation.
- Current methods struggle with muscle localization due to data limitations and model inaccuracies.
Purpose of the Study:
- To introduce MyoFormer, a structure-aware, multi-representation transformer.
- To enable muscle-level localization and classification of pathological gait.
- To overcome limitations of existing gait analysis techniques.
Main Methods:
- MyoFormer integrates joint coordinates and bone vectors from skeletal keypoints.
- Spatial and temporal attention mechanisms encode skeletal topology and gait periodicity.
- Temporal attention on vectors captures multi-scale features for comprehensive gait analysis.
Main Results:
- MyoFormer achieved 86.4% accuracy on synthetic data.
- Exploratory validation on a clinical cohort showed 61.7% Top-1 accuracy.
- The model demonstrates sensitivity to subtle, evolving gait abnormalities.
Conclusions:
- MyoFormer shows promise for precise pathological gait analysis.
- The model offers potential as an auxiliary diagnostic tool in clinical settings.
- This approach advances the field of gait disorder assessment.
