Detecting Differences in Gait Between Healthy individuals and Amputees Using Principal Component Analysis and Self-Organizing Maps

Abstract

This study presents an approach for analyzing and detecting variations in gait between healthy individuals and above-knee amputees with lower-limb prosthetics using principal component analysis (PCA) and self-organizing maps (SOM). The methodology begins with the extraction of principal components from the angular movements of the hip, knee, and ankle joints to capture the most significant movement patterns observed during walking in the sagittal plane. The SOM network then uses these principal components, or principal movements, as inputs. The role of the SOM is to classify the data and automatically discern differences between healthy individuals and amputees based on the principal movement elements. Through the classification results of the SOM network for the principal components, the study demonstrates the potential of using SOM to detect differences due to prosthetic limbs, including distinctive movement patterns in the extension and flexion patterns of the three lower extremity joints (ankle, knee, and hip). The findings suggest that employing the principal component analysis of gait with SOM technology can aid in constructing a diagnostic system that supports medical decision-making and uses the variance in principal movement elements for rapid identification through neural networks. Furthermore, this method could improve lower limb prosthetic design and rehabilitation programs to restore natural gait mechanisms in amputees.