Neural Networks-Based Approach to Solve Inverse Kinematics Problems for Medical Applications.

Area of Science:

• Biomechanics
• Machine Learning
• Medical Technology

Background:

• Motion capture technology is increasingly vital in medicine for analyzing complex human movements.
• Accurate and efficient analysis of motion capture data is crucial for clinical applications.

Purpose of the Study:

• To develop a novel algorithm integrating machine learning and biomechanics for rapid and robust motion capture data analysis.
• To accurately estimate human joint angles from motion capture data.

Main Methods:

• Comparison of multilayer perceptron and recurrent neural network (RNN) models for joint angle estimation.
• Pre-training neural networks using kinematic model data of the human arm with three degrees of freedom.
• Analysis of wrist and hand movements including flexion/extension, ulnar/radial deviation, and pronation/supination.

Main Results:

• A recurrent neural network with long short-term memory architecture demonstrated superior performance in solving the inverse kinematics problem for three rotational degrees of freedom.
• The RNN model achieved the lowest error and operated faster than real-time.
• The model's predictions exhibited robustness against noise in the motion capture data.

Conclusions:

• Pre-trained neural networks offer a feasible solution for real-time joint angle calculations from motion capture data.
• The developed algorithm provides a robust and efficient method for analyzing human movement in medical contexts.