Early Detection of Parkinson's Disease Using Deep Learning: A Convolutional Bi-Directional GRU Approach

Early Detection of Parkinson's Disease Using Deep Learning: A Convolutional Bi-Directional GRU Approach

Dhaya Chinnathambi, Srivel Ravi, Hariharan Dhanasekaran, Viswanathan Dhandapani, Ramana Rao, Saravanan Pandiaraj
Copyright: © 2024 |Pages: 13
DOI: 10.4018/979-8-3693-1115-8.ch013
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Machine learning is an evolving technology, which uses deep learning, a subset of artificial intelligence, harnesses neural networks, mirroring the human brain, for feature extraction and manipulation. Models like CNN, RNN, LSTM, SOM, among others, expedite data interpretation from vast datasets. Parkinson's disease, a neurodegenerative ailment primarily impacting dopamine-producing neurons within the substantia nigra of the brain, lacks a known cause and cure. Deep learning models play a pivotal role in early Parkinson's detection. In this study, a convolutional bi-directional GRU approach is employed to identify Parkinson's disease, with MAYFLY optimization for feature selection. Utilizing handwriting samples from Parkinson's patients, the proposed algorithm achieves a remarkable 96.40% accuracy in predicting the disease, facilitating early treatment interventions for affected individuals.
Chapter Preview
Top

Sarin et al. (2023) aimed to develop a three-step fuzzy classifier approach for dynamic handwriting analysis-based diagnosis of Parkinson's disease. The study used a dataset of 60 subjects, including 30 Parkinson's disease patients and 30 healthy individuals.

Dutta et al. (2023) involved the development of a new algorithm called Optimal Kernel Extreme Learning Machine with Colliding Bodies Optimization Algorithm (CBO-OKELM) for the identification and categorization of Parkinson's disease.

Zhao and Li (2023) sought to determine how PD patients' and healthy controls' handwriting differed using a hybrid model. Another study has explored the use of drawings for PD detection, using Convolutional Neural Networks (CNNs).

Bernardo et al. (2021) uses a support vector machine (SVM) as a classifier and the SqueezeNet convolutional neural network (CNN) model as a feature extractor to examine variations between hand-drawn images of people with Parkinson's disease (PD) and healthy individuals.

Parziale et al. (2021) claims that two neural network-based models, the Voice Impairment Classifier and the VGFR Spectrogram Detector, have been released with the goal of assisting medical professionals and patients in the early detection of sickness.

Johri and Tripathi (2019) aimed to compare different machine learning techniques for PD diagnosis based on handwriting samples. A study by Parziale et al. has explored the use of Cartesian genetic programming for the diagnosis of Parkinson's Disease (PD) through handwriting analysis.

de Souza et al. (2021) aimed to develop a computer-assisted diagnosis system for Parkinson's illness employing methods for restricted Boltzmann machines (RBM) and fuzzy optimum-path forest (FOPF). The study used a dataset of 200 subjects, including 100 Parkinson's disease patients and 100 healthy individuals.

Ali et al. (2019) is based on feature selection and adaptive boosting model, and it showed promising results in detecting PD. Another study has used transfer learning algorithm on a handwriting dataset for the detection of PD, which showed better performance.

Complete Chapter List

Search this Book:
Reset