Parkinson's Disease Detection Using R-CNN

Abstract

Pattern recognition in the biological sciences is significantly aided by machine learning techniques and
algorithms. These methods have helped academics categorise medical pictures and forecast models to have a
thorough grasp of challenging medical issues. In this study, region-based CNN (R-CNN) has been used to
separate or distinguish brains afflicted by Parkinson's disease (PD) from brains that are healthy and normal.
Complex clinical data must be categorised in order to identify disorders like Parkinson's disease or determine
the disease's stage. A machine learning system called R-CNN uses rich data sources including MRI, spiral and
wave drawings as datasets for the detection of Parkinson disease (PD) and prediction models for precise picture
categorization. This study's objective was to assess how well deep learning Information based on a quicker RCNN model is available to identify imaging features suggestive of idiopathic Parkinson's (PD). Parkinson's
disease (PD) affects neurological, behavioural, and physiological functions. Early Parkinson's disease changes
are small, making diagnosis challenging. Pathologists and neurologists assess PD patients' speech, writing,
walking, tremor, facial expressions, and drawing. Traditional machine learning techniques call for a number of
human processes, including decomposition, feature extraction, and classification. High accuracy was
demonstrated by faster R-CNN when separating PD from non-PD patterns. We effectively distinguished Spiral
and Wave of PD patients from healthy controls in MRI data using R-CNN, and we achieved accuracy of 100%
during training and 96-.99% during testing using batch normalisation. Our created prototype outperformed all
currently used state-of-the-art methods, and it is now prepared to be verified with more varied datasets in the
future.