Skip to main content
ISPRM 2026 Main banner
Final Program


Final Congress Guide

Biomedical Sciences

100 Section - Smart Lower-Limb Spasticity Assessment via Embeddable Pressure Insoles and Rotational Clustering Granular Logistic Regression

Wednesday, May 20, 2026
6:00 PM - 7:30 PM PT

    Presenting Author/Autor expositor(s)

  • wq

    wenting qin, MD

    Rehabilitation physician
    Shanghai Yangzhi Rehabilitation Hospital (Shanghai Sunshine Rehabilitation Center)
    shanghai, Shanghai, China (People's Republic)

    • Co-Author/Coautor(s)

  • qc

    qiangqiang chen, MD

    student
    the College of Biomedical Engineering, Fudan University
    shanghai, Shanghai, China (People's Republic)

  • LL

    Lingyu Liu, MD

    chief physician
    Shanghai YangZhi Rehabilitation Hospital (Shanghai Sunshine Rehabilitation Center)，School of Medicine，Tongji University
    shanghai, Shanghai, China (People's Republic)

    • Corresponding Author/Autor a quien dirigir la corr(s)

  • LJ

    Lingjing Jin, PhD

    president
    Yangzhi Rehabilitation Hospital (Shanghai Sunshine Rehabilitation Center)
    shanghai, Shanghai, China (People's Republic)

Objectives : Spasticity in lower limbs is a common motor dysfunction post-stroke, and its accurate assessment is essential for effective treatment and rehabilitation. However, the variability in muscle activation patterns complicates automated evaluation. This study aims to develop a convenient and precise system for automated spasticity assessment using flexible plantar pressure (PP) sensing insoles.

Design: We designed flexible, non-intrusive PP sensing insoles with 48 sensing units, integrated into regular shoes without external hardware. PP data were collected from 49 patients with lower limb spasticity. A total of 297 high-dimensional PP features were extracted to characterize spastic movement patterns. A novel rotational clustering granulation logistic regression (RCGLogR) algorithm was proposed to classify patients with plantar flexor spasticity (PFS) and combined plantar flexor-knee extensor spasticity (CPFKES).

Results: Using leave-one-subject-out cross-validation, RCGLogR achieved classification accuracies of 93.9% for PFS and 87.6% for CPFKES. It outperformed six classic machine learning algorithms by over 8% and two existing granular classifiers by more than 9% in accuracy for both patient types. The system demonstrated high diagnostic performance using only the smart insoles and the proposed algorithm.

Conclusion: This study presents the first automated system for precise lower limb spasticity assessment. Combining convenience with high accuracy, it addresses current limitations of manual evaluations reliant on clinician experience. The proposed approach shows strong potential for improving spasticity diagnosis and rehabilitation planning.