Classification of the Severity Level for Lower Limb Joint Injuries

This project is to study the classification of the severity level of lower limb joint injuries. Lower limb joint injuries can be defined as the limitations of hip, knee and ankle joint Range of Motion (ROM) during body movement. The medical measurement system, such as Electrogoniometer (EGM), Goniometer (GM) and Visual Tracking System (VTS) are commonly used to measure ROM in order to identify the severity level of lower limb joint injuries, which categorised into normal, mild, moderate and severe. There are 20 healthy subjects (10 male and 10 female) and 52 lower limb injury subjects are used in this study. In this project, the classification approach is introduced using the statistical analysis method for feature extraction. Thus, based from the statistical analysis results, it was discovered that mean, variance, standard deviation, standard error of mean, f-value, p-value, t-value and confidence interval for upper and lower boundary are the nine significant features. These extracted features are then classified using Artificial Neural Networks Levenberg-Marquardt (ANN-LM), Artificial Neural Networks Scaled Conjugate Gradient (ANN-SCG) and k-Nearest Neighbor (k-NN). From results attained, EGM is concluded as the best system for measurement of ROM and has proved that ANN-SCG has the highest classification accuracy of 94%, 95% and 90% for hip, knee and ankle compared to ANN-LM with accuracy of 68.8%, 85% and 82.5% and k-NN with accuracy of 89%, 83% and 87.64%.