Detection of Parkinson's Disease via Clifford Gradient-Based Recurrent Neural Network Using Multi-Dimensional Data

Disease prediction is a vital step in the early diagnosis of many diseases in the overpopulated modern world. The prediction has gotten simpler due to advancements in various machine learning (ML) techniques. However, the complexity of the model and the choice of the best machine-learning method for the given dataset significantly impact its accuracy. Globally, there are many datasets, but their unstructured nature prevents them from being used in any useful way. To extract anything valuable for use in the actual world, various strategies are therefore accessible. To evaluate the model, accuracy now serves as a key metric. This research proposes a novel Cliff-PD to detect Parkinson's disease using a Clifford gradient RNN classifier with MRI and EEG signals. Initially, the MRI images are denoised using multi-scale Retinex (MSR), and the EEG signal is denoised using Stationary Wavelet Transform filters to reduce the noise artifacts. Then, the Clifford Gradient RNN is employed to classify the normal, Non-specific white matter hyperintensity and global brain atrophy using MRI images. Furthermore, the Clifford Gradient RNN is employed to classify the normal, generalized background slowing using an EEG signal. The performance of the proposed Cliff-PD model achieved an accuracy of 99.18 %. Compared with SVMs, AlexNet's, and CROWD autoencoder, the accuracy range is improved overall by 5.38 %, 10.36 %, and 3.2 %, respectively.