Clinical Gait Analysis & Medical AI

Research Overview

My clinical motion-analysis work investigates how ordinary gait videos can provide quantitative and interpretable evidence of movement-related disorders. An important application is the assessment of adult spinal deformity (ASD), where a person’s gait can reveal functional characteristics that are not fully captured by static imaging.

I study spatiotemporal learning, phase-aware motion representations, and clinical knowledge-guided attention. These methods connect visual cues with domain knowledge—for example, emphasizing joints or body regions identified as diagnostically meaningful by clinicians—while evaluating performance on patient gait recordings.

Selected Contributions

• Two-stage video-based CNNs (2023): Spatiotemporal gait video classification for adult spinal deformity. Frontiers in Neuroscience
• PhaseMix (2024): Periodic motion fusion to improve gait-video classification. IEEE Access
• FilterNet (2026): Selection of phase-relevant frames for periodic motion analysis. ICEIC 2026
• Clinical knowledge-guided attention (2026): An anatomy-informed attention framework evaluated on monocular gait videos. SPIE Medical Imaging · Journal of Medical Imaging

These methods are intended as research tools for quantitative assessment and potential screening support, not replacements for clinical diagnosis.