Development of a quantitative gait analysis in an osteoarthritis rat model using machine learning

Area of Science:

• Biomedical Engineering
• Animal Models
• Machine Learning in Biology

Background:

• Quantitative gait analysis is crucial for evaluating osteoarthritis (OA) models.
• Current methods can be costly, invasive, or require markers.
• Developing accessible gait analysis tools is essential for OA research.

Purpose of the Study:

• To establish a low-cost, non-invasive, and markerless gait analysis pipeline for rat osteoarthritis models.
• To utilize machine learning for accurate classification of gait abnormalities.
• To provide a reproducible and unbiased method for movement disorder research.

Main Methods:

• Induction of rat osteoarthritis model via intraarticular injection of monosodium iodoacetate.
• Markerless gait analysis using DeepLabCut for joint coordinate extraction.
• Hierarchical clustering and principal coordinates analysis for data interpretation.

Main Results:

• Machine learning accurately separated osteoarthritis and healthy rat groups based on gait features.
• Principal coordinates analysis confirmed significant group separation while maintaining intra-group variability.
• The developed pipeline demonstrated high accuracy in classifying gait abnormalities.

Conclusions:

• A reproducible, non-biased, and cost-effective gait analysis method for OA rats was successfully developed.
• This markerless approach using machine learning is suitable for OA and other movement disorder models.
• The pipeline offers a valuable tool for quantitative assessment in preclinical research.