Assessment of Balance Control Subsystems by Artificial Intelligence

Area of Science:

• Biomechanics
• Artificial Intelligence
• Rehabilitation Science

Background:

• Current AI-based balance assessments are limited, often failing to evaluate individual balance control subsystems (BCSes).
• Balance control is complex, involving multiple subsystems that require precise, individual evaluation.

Purpose of the Study:

• To implement artificial intelligence (AI) for precise evaluation of different balance control subsystems (BCSes).
• To assess the feasibility of using AI to analyze center of pressure (CoP) data for detailed balance performance scoring.

Main Methods:

• Extracted 224 features from center of pressure (CoP) data for each participant.
• Employed regressors to map extracted features to physical therapist scores for Mini-BESTest and its BCS sub-scores (APA, RPC, SO, DG).

Main Results:

• AI achieved minimum mean absolute errors of 2.658 for total Mini-BESTest score, 0.827 for anticipatory postural adjustments (APA), 0.970 for reactive postural control (RPC), 0.642 for sensory orientation (SO), and 0.98 for dynamic gait (DG).

Conclusions:

• This preliminary study demonstrates the potential of AI for precise assessment of individual balance control subsystems (BCSes).
• The developed AI approach shows promise for future clinical applications in balance evaluation.