Gait quality assessment using self-organising artificial neural networks

Area of Science:

• Biomechanics
• Artificial Intelligence
• Medical Diagnostics

Background:

• Gait analysis is crucial for diagnosing and monitoring conditions affecting movement.
• Current methods may not fully capture the complexity of gait deviations.
• Advanced computational techniques offer potential for more precise gait assessment.

Purpose of the Study:

• To develop and validate a novel method for quantifying gait deviation using self-organising neural networks.
• To identify specific causes of abnormal gait patterns in patients.
• To enhance the diagnostic utility of gait analysis in clinical settings.

Main Methods:

• Utilized three-dimensional joint angles, moments, and powers of the lower limbs and pelvis.
• Trained Kohonen artificial neural networks to define a 'normal' gait pattern.
• Quantified gait quality using the quantisation error derived from patient data.

Main Results:

• The artificial neural network successfully learned an abstract representation of normal gait.
• Quantisation error effectively measured gait quality and deviation.
• Sensitivity analysis pinpointed anatomical locations and timing of gait abnormalities.

Conclusions:

• Self-organising neural networks provide a robust method for quantifying gait deviation.
• The quantisation error serves as an advanced gait index, offering insights into the causes of deviations.
• This approach enhances the potential of gait analysis for clinical diagnosis and patient management.