Wearable multimodal sensors for evaluation of patients with Parkinson disease

For a population that is moving towards an elderly stage of development, Parkinson's disease (PD) is characterized in the second place for the most common chronic progressive neurodegenerative illness in the world after Alzheimer's disease, which regularly affects older generation. In the next 30 years, this amount is estimated to double due to the increase in the number of ageing people, as age is the leading key risk feature for the start of PD. There are a variety of medications, such as levodopa available to treat PD. With the latest advancement in healthcare technology, current researches permit the monitoring of PD with the application of wearable sensor technology. From previous studies, researchers have realized the application of wearable sensors as a useful tool that had the capability to differentiate various types of PD symptoms using uni-modal sensor or bi-modal sensors (accelerometer and gyroscope). Therefore, early diagnosis of PD through multimodal wearable technology can be considered for this aim. In this paper, the data are collected using on-body triaxial wearable sensors (accelerometer, gyroscope and magnetometer) for classifying people with Parkinson (PWP) from healthy controls. The system performance was characterized based on 10-fold cross validation method, applying the proposed time and frequency domain features and classification algorithms. The strength of the proposed method has been evaluated through several performance measures. In summary, these results show that the proposed machine learning techniques had ability in differentiating PWP from healthy controls with highest average accuracy, sensitivity, specificity and ROC of above 88%.