DAGAN-Based Gait Features Augmentation for Ankle Instability Detection.

Area of Science:

• Biomechanics
• Medical Imaging
• Artificial Intelligence

Background:

• Ankle ligament injuries pose diagnostic challenges.
• Accurate gait analysis is crucial for diagnosis.
• Limited medical data hinders AI model development.

Purpose of the Study:

• To develop a computer-aided diagnostic program for ankle ligament injuries.
• To address the challenge of insufficient medical samples using data augmentation.
• To evaluate the impact of enhanced gait data on diagnostic model performance.

Main Methods:

• Utilized the Heidelberg Foot Measurement Method (HFMM) for gait kinematic data.
• Proposed an Attention Dual Generative Adversarial Network (Attention-DualGAN) for gait feature augmentation.
• Employed Convolutional Long Short-Term Memory (ConvLSTM) and Time Convolution Network (TCN) models for detection.

Main Results:

• Attention-DualGAN generated high-quality synthetic gait kinematics.
• ConvLSTM model accuracy improved by 2.4% post-data enhancement.
• TCN model accuracy improved by 7.47% post-data enhancement.
• Synthetic data showed high correlation with real gait data.

Conclusions:

• Attention-DualGAN effectively augments gait feature datasets.
• Enhanced datasets significantly improve ankle ligament injury diagnostic model performance.
• The study demonstrates the potential of AI-driven data augmentation in medical diagnostics.