Major depressive disorder (MDD) contributes the most to human’s functional frailty worldwide. Therefore, its timely diagnosis and treatment is of utmost importance. Conventionally, MDD is diagnosed using subjective evaluation methods, so, it is essential to develop a quantitative biomarker for its automated diagnosis. Accordingly, this study proposes a 2D-CNN network and a new biomarker for automated detection of MDD. The proposed biomarker is developed by estimating wavelet coherence (WCOH) amongst the brain’s default mode network (DMN) regions using EEG signals. This biomarker data from 30 MDD patients and 30 healthy controls (HCs) is randomly divided into training and testing sets for network training and blind testing, respectively. The performance of the network is evaluated via 10-fold cross-validation which is applied to the training data only to avoid learning bias. The blind testing of subjects is performed using two different classification approaches i.e., sample-based and subject-based. The former achieves 98.1% accuracy, 98.0% sensitivity, and 98.2% specificity whereas the latter yields 100% each for accuracy, sensitivity, and specificity. This high classification performance validates that DMN-based WCOH can be used as a potential biomarker and that the proposed 2D-CNN can provide reliable performance assessment for the diagnosis of MDD.

Deep metric learning (DML) has achieved state-of-the-art results in several deep learning applications. However, this type of deep learning models has not been tested on the classification of electrical brain waves (EEG) for brain computer interface (BCI) applications. For the first time, we propose a triplet network to classify motor imagery (MI) EEG signals. Stockwell Transform has been used for converting the EEG signals in the time domain into the frequency domain, which resulted in improved DML classification accuracy in comparison to DML with Short Term Fourier Transform (0.647 vs. 0.431). DML model was trained with a topogram of concatenated 64 EEG channel spectrograms. The training batch was comprised of triplet pairs of the anchor, positive, and negative labeled epochs. The triplet network was able to train an embedding feature space that minimized the Euclidean distance between the embeddings of spectrograms of the same class and increased the distance between the embeddings of different labeled images. The proposed method has been tested on an EEG dataset of 109 untrained subjects. We showed that the DML classifier is able to converge with an extremely small number of training samples (~120 EEG trials) for only one subject per model, mitigating the well-known issue of the large inter-individual variability of human MI-BCI EEG which degrades the classification performance. The proposed preprocessing pipeline and the Triplet Network provide a promising method to classify MI-BCI EEG signals with much less training samples than the previous methods.

This study aims to investigate the effects of workplace noise on neural activity and alpha asymmetries of the prefrontal cortex (PFC) during mental stress conditions. Workplace noise exposure is a pervasive environmental pollutant and is negatively linked to cognitive effects and selective attention. Generally, the stress theory is assumed to underlie the impact of noise on health. Evidence for the impacts of workplace noise on mental stress is lacking. Fifteen healthy volunteer subjects performed the Montreal imaging stress task in quiet and noisy workplaces while their brain activity was recorded using electroencephalography. The salivary alpha-amylase (sAA) was measured before and immediately after each tested workplace to evaluate the stress level. The results showed a decrease in alpha rhythms, or an increase in cortical activity, of the PFC for all participants at the noisy workplace. Further analysis of alpha asymmetry revealed a greater significant relative right frontal activation of the noisy workplace group at electrode pairs F4-F3 but not F8-F7. Furthermore, a significant increase in sAA activity was observed in all participants at the noisy workplace, demonstrating the presence of stress. The findings provide critical information on the effects of workplace noise-related stress that might be neglected during mental stress evaluations.

This research proposes a new hybrid algorithm for automatic removal of eye blink artifact from EEG data based on empirical mode decomposition (EMD) and canonical correlation analysis (CCA). The validity and efficiency of the proposed algorithm is evaluated using correlation coefficient and signal-to-artifact ratio (SAR) and the proposed algorithm is also compared with other popular eye blink artifact removal techniques (CCA, ICA, EMD-ICA) on simulated EEG data of two channels. From the simulation results, the average correlation coefficients for the EEG channels are obtained as 0.908 and 0.864 respectively. The SAR of the EEG signal also improved from 2.2 dB to 6.0 dB after correction using our proposed method. Compared to other eye blink artifact removal techniques, our proposed method has two benefits. Firstly, no visual inspection is required to detect the eye blink artifact components. Secondly, computational assessment of corrected EEG waveforms reveals that the proposed algorithm retrieves the EEG data by removing the eye blink artifacts reliably.

Morphological changes related to different diseases that occur in the retina are currently extensively researched. Manual segmentation of retinal fluids is time-consuming and subject to variability, giving prominence to the demand for robust automatic segmentation methods. The standard in assessing the existence and mass of retinal fluids at present is through the optical coherence tomography (OCT) modality. In this study, semantic segmentation deep learning networks were examined in 2.5D and ensembled with 2D networks. This analysis aims to show how these networks can perform in-depth than using only a single B-scan and the effects of 2.5 patches when fitted to the deep networks. All experiments were evaluated using public data from the RETOUCH challenge as well as the OPTIMA challenge dataset and Duke dataset. The networks trained in 2.5D performed slightly better than 2D networks in all datasets. The average performance of the best network was 0.867, using the dice similarity coefficient score (DSC) metric on the RETOUCH dataset. On the DUKE dataset, Deeplabv3+Pa outperformed other networks in this study with a dice score of 0.80. Experiments showed a more robust performance when networks were ensembled. Intraretinal fluid (IRF) was recognized better than other fluids with a DSC of 0.924. Deeplabv3+Pa model outperformed all other networks with a p-value average of 0.03 on the RETOUCH challenge dataset. Methods used in this study to distinguish retinal disorders outperform human performance as well as showed competitive results to the teams who joined both challenges. Three consecutive B-scans, including partial depth information in training neural networks, were stacked as a single image built for more robust networks compared to providing only 2D information.