Support vector machines for detecting age-related changes in running kinematics.

Area of Science:

• Biomechanics
• Data Mining
• Gerontology

Background:

• Classical inferential statistics struggle with numerous biomechanical gait variables.
• Previous studies often report non-significant age-related differences in running kinematics.
• Data mining offers a more generalized approach to analyze complex biomechanical data.

Purpose of the Study:

• To re-analyze lower extremity running kinematic data from young and elderly male runners.
• To apply the Support Vector Machine (SVM) classification approach to identify age-related differences.
• To assess the effectiveness of SVM in distinguishing running gait patterns between age groups.

Main Methods:

• Lower extremity running kinematic data from 17 young and 17 elderly male runners were analyzed.
• Support Vector Machine (SVM) classification was employed using 31 kinematic variables.
• Three kernel methods (linear, polynomial, radial basis function) were evaluated.
• A forward feature selection algorithm was used to identify key discriminating variables.

Main Results:

• The SVM classification approach yielded varying accuracy rates depending on the kernel method used.
• The linear kernel method demonstrated the highest classification performance.
• A subset of six kinematic features achieved 100% classification accuracy with the linear kernel SVM.
• These six features effectively captured age-related differences in running biomechanics.

Conclusions:

• The Support Vector Machine (SVM) approach shows significant potential for analyzing age-related differences in running gait biomechanics.
• A small set of kinematic features can powerfully distinguish between young and elderly runners.
• The SVM method is encouraged for application in clinical contexts for gait analysis.