Related Experiment Video
Updated: Jun 5, 2026
07:33
In Vitro Application of a Wireless Sensor in Flexion-Extension Gap Balance of Unicompartmental Knee Arthroplasty
Published on: May 5, 2023
A mechatronic and artificial intelligence-driven framework for automated non-invasive knee abnormality screening
Vidyapati Kumar1, Muhamed Shijas1, Hrishikesh M V1
1Department of Mechanical Engineering, Indian Institute of Technology, Kharagpur, West Bengal, India.
Summary
This study introduces a low-cost, automated screening tool for knee abnormalities using surface electromyography (sEMG) and goniometers. It achieves high accuracy, offering a more accessible alternative to MRI for detecting movement impairments.
Area of Science:
- Biomedical Engineering
- Machine Learning in Healthcare
- Biomechanics
Background:
- Current knee abnormality detection methods like MRI are expensive and inaccessible.
- Subjective clinical evaluations lack objectivity and consistency.
- There is a need for cost-effective, automated diagnostic tools for knee conditions.
Purpose of the Study:
- To develop an integrated mechatronic and machine learning framework for knee abnormality detection.
- To utilize surface electromyography (sEMG) and goniometers for multimodal mobility data acquisition.
- To introduce novel time-frequency features for improved classification accuracy.
Main Methods:
- Collected multimodal mobility data using sEMG sensors and goniometers.
- Engineered novel time-frequency features: Enhanced Mean Absolute Value (EMAV) and Enhanced Wavelength (EWL).
- Employed the Extra Trees classifier, optimizing with novel features and validated using Friedman and Nemenyi tests. Incorporated SHAP for interpretability.
Main Results:
- The Extra Trees classifier achieved a cross-validated accuracy of 94.7%.
- Novel features (EMAV, EWL) enhanced classifier performance by 3.16% compared to conventional methods (MAV, WL).
- Achieved 95% precision and recall, demonstrating robust diagnostic capability.
Conclusions:
- The developed framework offers a low-cost, automated solution for knee abnormality screening.
- This approach enhances accessibility and objectivity in diagnosing knee conditions.
- The framework is extendable to broader human movement analysis and pathology detection applications.