Parkinson’s Disease Detection Through Deep Learning Model

Early detection of Parkinson’s disease is imperative for efficient treatment and managing of the disease in a patient. The detection of the disease is not perfect as until now, it is having mostly been done manually by the humans. Since human observation is prone to mistakes and can generate inconsistency, there is a need to create a program that is able to improve upon the current diagnosis and detection of the disease with better accuracy. To make such a program, for this paper, a detection model is proposed that utilizes SPECT images of healthy and diseased persons to find the affected region of the brain, this region helps determine the region of interest (ROI) for the study. The model uses artificial neural network and image processing to accomplish its main task of detection. The area values of the ROI are then fed into the model which aims to calculate the pattern of the diseased region to determine the extent of Parkinson’s disease, and whether the brain is healthy or not. This simple but efficient ANN model classifies the subjects with and without PD, with an accuracy of 93%, sensitivity of 100%, and specificity of 86%. The ANN model undergoes constant improvement and fed larger datasets over time to improve the accuracy of the model and achieve a better performance than any previously available modes of detection. Hence, the proposed system has the potential to aid the doctors and nurses in the medical field of neurobiology and can be practically implemented to improve the diagnosis of Parkinson’s disease.