Abstract

Knee diseases such as osteoarthritis and patellofemoral disorder can cause abnormal patellar motion and lead to the reduction of a person’s activity ability. In this work, a new kind of flexible sensors combining graphene particles, lead magnesium niobate-lead titanate (PMN-PT) and polyvinylidene fluoride (PVDF) to form a flexible ternary composite is developed. A wearable device with six flexible sensors is specifically constructed to monitor the movement of patella. The position of sensors is designed based on medical mechanism of the patella movements. Testing data of patella movement is acquired from volunteers with and without knee injury. The raw signals are filtered to reduce high-frequency noises. Several features are extracted to characterize the motion of patella, including the number of change points and auto-correlation coefficient. The extracted features and the data are labeled according to Lysholm Knee Scoring Scale. Because the number of unhealthy data is much smaller than that of the healthy data, an oversampling strategy is applied to introduce new artificial data for subsequent machine learning classifier. Three classification methods including the Tree BAGGER, the k-nearest neighbors and the support vector machine are employed. Results show that the best performance is achieved by the support vector machine method with 96% accuracy.

This content is only available via PDF.
You do not currently have access to this content.