Resting-state electroencephalography based deep-learning for the detection of Parkinson’s disease

Abstract

Parkinson’s disease (PD) is one of the most serious and challenging neurodegenerative disorders to diagnose. Clinical diagnosis on observing motor symptoms is the gold standard, yet by this point nerve cells are degenerated resulting in a lower efficacy of therapeutic treatments. In this study, we introduce a deep-learning approach based on a recently-proposed 20-Layer Convolutional Neural Network (CNN) applied on the visual realization of the Wavelet domain of a resting-state EEG. The proposed approach was able to efficiently and accurately detect PD as well as distinguish subjects with PD on medications from subjects who are off medication. The gradient-weighted class activation mapping (Grad-CAM) was used to visualize the features based on which the approach provided the predictions. A significantly high accuracy, sensitivity, specificity, AUC, and Weighted Kappa Score up to 99.9% were achieved and the visualization of the regions in the Wavelet images that contributed to the deep-learning approach decisions was provided. The proposed framework can then serve as an effective computer-aided diagnostic tool that will support physicians and scientists in further understanding the nature of PD and providing an objective and confident opinion regarding the clinical diagnosis of the disease.

Parkinson’s disease detection based on multi-pattern analysis and multi-scale convolutional neural networks

“…E.g., our accuracy is about 14.12% higher than the linear-predictive-coding EEG algorithm proposed by Anjum et al (2020) , and about 10.52% higher than the Hjorth parameter and the gradient boosting decision tree algorithm proposed by Lee et al (2022) . For the UC San Diego dataset (15 PD patients and 16 controls), the classification performance (accuracy, sensitivity, specificity, and AUC are above 93% in γ band) of our proposed MCNN model is comparable, although not the highest, compared to other related studies ( Khare et al, 2021a , b ; Loh et al, 2021 ; Shaban, 2021 ; Shaban and Amara, 2022 ). Some previous work can achieve close to 100% accuracy ( Barua et al, 2021 ; Khare et al, 2021b ).…”

“… Barua et al (2021) achieved 99.93% and 100% classification accuracy for HC vs. PD_OFF and HC vs. PD_ON using the proposed novel aspirin pattern, respectively. Shaban and Amara (2022) used a continuous wavelet-based deep learning method to obtain a high accuracy of 99.9% for PD detection. Although these studies achieved slightly higher accuracy than our proposed model, their methods were only validated on a single dataset and did not generalize to other datasets.…”

“… Shaban (2021) used an artificial neural network-based framework and three EEG spatial channels to screen subjects as PD patients and HCs with 98% accuracy, 97% sensitivity, and 100% specificity. More recently, they recently introduced a 20-layer CNN-based deep learning method to identify HC, PD_OFF, and PD_ON using the wavelet domain of resting-state EEG with 99.9% accuracy ( Shaban and Amara, 2022 ).…”

Thin-Film Electrodes Based on Two-Dimensional Nanomaterials for Neural Interfaces

“…On the other hand, external neural interfaces of the EEG type are associated with a much lower risk for the user and can be used to noninvasively monitor brain states (including cognitive stress load and sleep patterns) and brain health, opening the possibility, for example, to diagnose epilepsy, as well as potentially screen for and achieve an early diagnosis of some of the neurodegenerative diseases. − In addition, noninvasive neural interfaces open the path toward all nonclinical hands-free control of autonomous external devices by harnessing the brain signals, including robotics, bionic prosthetics, neurogaming, consumer electronics, as well as autonomous vehicles. − The external devices can be controlled by the interpretation of human intention via detecting the cortical electrical activity, eye movement, and muscle movement through EEG, electrooculography (EOG), and electromyography (EMG), respectively, which all rely on noninvasive sensors. ,, The foundation concept of interconnecting external devices directly with brain signals was demonstrated by Hans Berger in 1929 by developing the first-generation EEG device using two channels with metallic needle electrodes to enable the noninvasive recording of neuroelectrical brain signals . The first scientific study on volitional control of human brain oscillation was reported by Kamiya et al in 1969, where the change of “alpha waves” (neural oscillations in the frequency range of 8–13 Hz) on continuous sensory feedback using EEG channels at the C 4 -O 2 , C z -A 2 positions had been demonstrated .…”

Generalizable electroencephalographic classification of Parkinson’s Disease using deep learning

“…Many of these studies report very high accuracies above 98%. [5 - 12] However, the methods employed in these articles may suffer from the limitation of information-leakage between the training and testing data [6]. This limitation can lead to large overestimation of classification ability on unseen, real-world data.…”