The novel quantitative technique for assessment of gait symmetry using advanced statistical learning algorithm

Area of Science:

• Biomechanics
• Clinical Biomechanics
• Statistical Learning

Background:

• Accurate gait asymmetry identification is crucial for assessing at-risk gait in clinical settings.
• Traditional methods may not capture subtle dynamic changes in gait patterns.
• Understanding gait dynamics is key to preventing falls and mobility issues.

Purpose of the Study:

• To investigate a classification method based on statistical learning to quantify gait symmetry.
• To assess the ability of this method to detect subtle differences between lower limb gait patterns.
• To compare the proposed method with traditional symmetry index methods.

Main Methods:

• Kinetic gait data from 60 participants were collected using a strain gauge force platform during normal walking.
• A classification method employing statistical learning algorithms, specifically support vector machine (SVM), was developed for binary classification.
• The algorithm was designed to quantitatively evaluate gait symmetry by analyzing statistical distributions of gait variables.

Main Results:

• The proposed statistical learning method effectively captured intrinsic dynamic information in gait variables.
• It demonstrated superior generalization performance in recognizing right-left gait patterns.
• The technique successfully identified small, significant differences between lower limbs, outperforming the traditional symmetry index method.

Conclusions:

• The developed classification method provides a more sensitive tool for quantifying gait symmetry.
• This approach can identify subtle gait asymmetries, particularly beneficial for early detection in the elderly.
• The algorithm shows potential as an effective tool for clinical diagnosis and gait analysis.