scholarly journals Deep Transfer Learning for Parkinson’s Disease Monitoring by Image-Based Representation of Resting-State EEG Using Directional Connectivity

Algorithms ◽  
2021 ◽  
Vol 15 (1) ◽  
pp. 5
Author(s):  
Emad Arasteh Emamzadeh-Hashemi ◽  
Ailar Mahdizadeh ◽  
Maryam S. Mirian ◽  
Soojin Lee ◽  
Martin J. McKeown

Parkinson’s disease (PD) is characterized by abnormal brain oscillations that can change rapidly. Tracking neural alternations with high temporal resolution electrophysiological monitoring methods such as EEG can lead to valuable information about alterations observed in PD. Concomitantly, there have been advances in the high-accuracy performance of deep neural networks (DNNs) using few-patient data. In this study, we propose a method to transform resting-state EEG data into a deep latent space to classify PD subjects from healthy cases. We first used a general orthogonalized directed coherence (gOPDC) method to compute directional connectivity (DC) between all pairwise EEG channels in four frequency bands (Theta, Alpha, Beta, and Gamma) and then converted the DC maps into 2D images. We then used the VGG-16 architecture (trained on the ImageNet dataset) as our pre-trained model, enlisted weights of convolutional layers as initial weights, and fine-tuned all layer weights with our data. After training, the classification achieved 99.62% accuracy, 100% precision, 99.17% recall, 0.9958 F1 score, and 0.9958 AUC averaged for 10 random repetitions of training/evaluating on the proposed deep transfer learning (DTL) network. Using the latent features learned by the network and employing LASSO regression, we found that latent features (as opposed to the raw DC values) were significantly correlated with five clinical indices routinely measured: left and right finger tapping, left and right tremor, and body bradykinesia. Our results demonstrate the power of transfer learning and latent space derivation for the development of oscillatory biomarkers in PD.

2021 ◽  
Vol 7 (1) ◽  
Author(s):  
Nicholas D’Cruz ◽  
Griet Vervoort ◽  
Sima Chalavi ◽  
Bauke W. Dijkstra ◽  
Moran Gilat ◽  
...  

AbstractThe onset of freezing of gait (FOG) in Parkinson’s disease (PD) is a critical milestone, marked by a higher risk of falls and reduced quality of life. FOG is associated with alterations in subcortical neural circuits, yet no study has assessed whether subcortical morphology can predict the onset of clinical FOG. In this prospective multimodal neuroimaging cohort study, we performed vertex-based analysis of grey matter morphology in fifty-seven individuals with PD at study entry and two years later. We also explored the behavioral correlates and resting-state functional connectivity related to these local volume differences. At study entry, we found that freezers (N = 12) and persons who developed FOG during the course of the study (converters) (N = 9) showed local inflations in bilateral thalamus in contrast to persons who did not (non-converters) (N = 36). Longitudinally, converters (N = 7) also showed local inflation in the left thalamus, as compared to non-converters (N = 36). A model including sex, daily levodopa equivalent dose, and local thalamic inflation predicted conversion with good accuracy (AUC: 0.87, sensitivity: 88.9%, specificity: 77.8%). Exploratory analyses showed that local thalamic inflations were associated with larger medial thalamic sub-nuclei volumes and better cognitive performance. Resting-state analyses further revealed that converters had stronger thalamo-cortical coupling with limbic and cognitive regions pre-conversion, with a marked reduction in coupling over the two years. Finally, validation using the PPMI cohort suggested FOG-specific non-linear evolution of thalamic local volume. These findings provide markers of, and deeper insights into conversion to FOG, which may foster earlier intervention and better mobility for persons with PD.


PLoS ONE ◽  
2017 ◽  
Vol 12 (3) ◽  
pp. e0172394 ◽  
Author(s):  
Robert Westphal ◽  
Camilla Simmons ◽  
Michel B. Mesquita ◽  
Tobias C. Wood ◽  
Steve C. R. Williams ◽  
...  

2016 ◽  
Vol 22 ◽  
pp. e156
Author(s):  
Sun Nee Tan ◽  
Yiming Zhang ◽  
Aiping Liu ◽  
Jane Wang ◽  
Martin J. McKeown

Sign in / Sign up

Export Citation Format

Share Document