A neural network system for diagnosis and assessment of tremor in parkinson disease patients

Tremor is one of the most important symptom in Parkinson´s disease, which has been assessed clinically by neurologists as part of UPDRS scale. In this paper, we have implemented a supervised learning pattern recognition system to assess UPDRS of each Parkinson patient tremor to fill the absence of a reliable diagnosis and monitoring system for Parkinson patients. In our system a simple noninvasive method based on the recorded acceleration through the smartphone have been used for data acquisition. The results show high accuracy in the classifier block and neural network. A tight correlation between UPDRS scale and acceleration values reveals 91 percent accuracy by neural network with two hidden layers.