Automatic Diagnosis of Cerebral Palsy Gait Using Computational Intelligence Techniques: A Low-Cost Multi-Sensor...

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Gait Analysis

Background:

• Quantitative gait assessment is vital for evaluating Cerebral Palsy (CP) interventions.
• Current systems are often costly and require significant user input.
• There is a need for accessible and automated CP gait diagnostic tools.

Purpose of the Study:

• To develop and validate a low-cost, automated system for diagnosing Cerebral Palsy (CP) gait.
• To investigate the effectiveness of spatio-temporal gait features, particularly the speed-invariant walking ratio, for CP gait detection.
• To compare the performance of various classifiers, including Extreme Learning Machine, for CP gait diagnosis.

Main Methods:

• A low-cost gait assessment system utilizing multiple Kinect sensors was designed.
• A data-driven algorithm was developed to remove outlier frames from multi-sensor gait data.
• Spatio-temporal features, including the walking ratio, were extracted and analyzed.
• Supervised classifiers and Extreme Learning Machine were employed to diagnose CP gait.

Main Results:

• The proposed system achieved high diagnostic accuracy (≈98%), with 100% sensitivity and 96.87% specificity.
• Outlier removal significantly improved abnormality detection performance.
• The walking ratio was identified as the most effective feature for CP gait detection.
• Classifier performance notably increased when using the walking ratio, with Extreme Learning Machine showing competitive results.

Conclusions:

• The developed low-cost system offers an effective and accurate solution for automated CP gait diagnosis.
• The walking ratio is a powerful, speed-invariant feature for identifying CP gait abnormalities.
• This system presents a promising, accessible tool for quantitative gait assessment in CP patients.