Parkinson's disease Diagnosis using a Combined Deep Learning Approach

Parkinson's disease, which often affects the elderly, is one of the most dangerous neurodegenerative disorders. The prime characteristics observed in people with this disorder are: bradykinesia i.e. slowness of movement, rigidity, tremor and instable posture. One of the most commonly observed changes is in the handwriting of a person with Parkinson's disease. Due to the tremor effects on the hands, drawing a spiral shape in particular becomes difficult for those with Parkinson's. In this work, we have considered this as the primary criteria for our diagnosis system. Deep learning is an enhancement of Machine learning and has presently demonstrated efficiency to analyze and diagnose unstructured datasets and, hence to provide the best care for patients. In this paper, an early diagnosis technique combining LeNet and Long Short Term Memory (LSTM) has been proposed for Parkinson's disease.