Detecting Early Parkinson’s Disease via Computer Keyboard Interaction by Using Machine Learning
One of the most popular progressive neuro-degenerative motor diseases in the world that has influence on more than 6.3 million people is Parkinson’s disease (PD). Currently there is no reliable test for detecting PD by non-specialist physicians, particularly in the early stage of the disease in which the signs might be poor and subtly characterized. This leads to misdiagnose the disease by non-physicians up to 25%. In both motor and non-motor symptoms lack of neurons that produce dopamine will occur and individuals might be the holder of the disease for couple of years before diagnosis. There is a high demand for an objective and more accurate of detecting early PD, especially the ones that could be used personally at home or office. Timing information of keystroke from 85 individuals (that consist of 42 with PD and 43 controls) were taken in this investigation as they have been typed on a keyboard of computer with an extended period and displayed that PD has impact on many characteristics of finger and hand movements. When this done to two participants groups, our proposed bagging ensemble model were capable to differentiate the early PD from the others successfully with 82% accuracy and 0.86 AUC. The approach does not depend on the skill or experience of the specialist and does not need any medical supervision and specialized equipment. For more general applications, PD cannot be distinguished from similar related disorders because the symptoms of the second underlying disorder are not currently incorporated.