Sarcopenia classification model for musculoskeletal patients using smart insole and artificial intelligence gait...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Gerontology

Background:

• Physical function, musculoskeletal disorders, and sarcopenia have a complex relationship.
• Current physical function tests (e.g., gait speed, chair stand) are subjective.
• Smart devices with inertial measurement unit sensors and AI are emerging to overcome these limitations.

Purpose of the Study:

• To develop and evaluate AI-based classification models for sarcopenia detection.
• To compare the efficacy of smart insole and pose estimation technologies.
• To explore the integration of these technologies for improved diagnostic accuracy.

Main Methods:

• Utilized smart insole devices and AI-based pose estimation.
• Employed Random Forest (RF), Support Vector Machine, and Artificial Neural Network models.
• Classified 83 individuals into control and sarcopenia groups using training and testing datasets.

Main Results:

• Pose estimation alone achieved high accuracy (0.92-0.96) and F1-scores (0.94-0.97).
• Key variables identified by RF included 'Hip_dif', 'Ankle_dif', and 'Hipankle_dif'.
• Combining smart insole and pose estimation data further improved accuracy (0.80-1.00) and F1-scores (0.73-1.00).

Conclusions:

• An integrated classification model using smart insole and pose estimation shows significant potential.
• The RF model demonstrated strong performance, especially with 'Hip' and 'Ankle' variables.
• Advanced digital biomarkers and AI integration promise enhanced diagnostics and interventions for conditions like sarcopenia.