Empowering Lower Limb Disorder Identification through PoseNet and Artificial Intelligence

Area of Science:

• Biomechanics
• Medical Imaging
• Machine Learning

Background:

• Lower limb disorders affect mobility and quality of life.
• Accurate diagnosis is crucial for effective treatment planning.
• Current diagnostic methods can be invasive or lack detailed kinematic information.

Purpose of the Study:

• To develop and evaluate a novel, non-invasive method for classifying lower limb disorders.
• To utilize gait analysis and PoseNet features for identifying knee, hip, and ankle conditions.
• To compare the performance of various machine learning algorithms for this classification task.

Main Methods:

• Gait analysis using video data and the PoseNet algorithm to extract key joint movements.
• Standardization of extracted features for input into machine learning models.
• Training and testing of Random Forest, Extra Tree Classifier, Multilayer Perceptron, Artificial Neural Networks (ANN), and Convolutional Neural Networks (CNN) using K-fold cross-validation on a dataset of 174 patients.

Main Results:

• The study achieved high accuracy and precision in classifying lower limb disorders.
• Artificial Neural Networks (ANN) demonstrated the highest classification accuracy at 98.84%.
• The proposed method proved effective in differentiating various lower limb conditions.

Conclusions:

• The developed methodology offers a non-invasive and efficient approach for diagnosing lower limb disorders.
• PoseNet-based gait analysis combined with machine learning shows significant potential for improving diagnostic accuracy.
• This approach can aid in better treatment planning for patients with knee, hip, and ankle conditions.