Depression recognition based on the reconstruction of phase space of EEG signals and geometrical features

Akbari, Hesam; Sadiq, Muhammad Tariq; Rehman, Ateeq Ur; Ghazvini, Mahdieh; Payan, Malih; Bagheri, Hourieh; Bagheri, Hamed

doi:10.1016/j.apacoust.2021.108078

Cited by 63 publications

(30 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, it is difficult to partition the available EEG dataset of small sample size into a training dataset used for training classifiers and an independent test dataset of sufficient size. Due to this limitation, previous studies have only estimated the efficacy of their method in MDD detection by performing a k-fold CV on the entire EEG dataset, where k = 10 [10][11][12][13][14][15][16]18,19,22,25] or k = number of EEG samples [9,17,23] (i.e., leave-one-out (LOO)). Directly performing a 10-fold CV on the group of participants (i.e., partitioning the entire group into 10 folds, with no EEGs of the same participants appearing in different folds at the same time) is nearly impossible due to participant size limitations.…”

Section: Cross Validation Used In Previous Work: More Detailed Review and Analysismentioning

confidence: 99%

“…Many studies have tried to apply machine learning (ML) methods to classify resting-state (e.g., [ 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 ]) versus task-related (e.g., [ 21 , 24 , 25 ]) electroencephalography (EEG) signals between MDD and healthy control (HC) groups. A few studies have also proposed multi-model systems that fuse EEG signals and other physiological data (e.g., [ 26 ]).…”

Section: Introductionmentioning

confidence: 99%

“…The sample sizes in these ML-based studies [ 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 ] were all less than 100 (24~90), except for two studies by Cai et al [ 12 , 15 ] in which the subject numbers were 267 and 213, respectively (however, they only measured a very low-density EEGs from three electrodes at FP1, FP2, and Fz). Previous studies, due to limited data size, have only performed cross validation procedures on limited EEG data to estimate the generalization performance of their proposed algorithms/models, without using an independent test set of sufficiently large size.…”

Section: Introductionmentioning

confidence: 99%

“…These inevitable factors of EEG variability have been considered to be a challenge to real-world brain-computer interface (BCI) [ 29 , 30 ] and EEG-based applications (e.g., seizure detection [ 31 ]). In this view, most of the previous EEG studies [ 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ] have lacked a validation on datasets with high diversity (or large EEG variability) because their data were all collected from a single site.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Resting-State EEG Signal for Major Depressive Disorder Detection: A Systematic Validation on a Large and Diverse Dataset

Huang²,

Huang³

et al. 2021

Biosensors

View full text Add to dashboard Cite

Major depressive disorder (MDD) is a global healthcare issue and one of the leading causes of disability. Machine learning combined with non-invasive electroencephalography (EEG) has recently been shown to have the potential to diagnose MDD. However, most of these studies analyzed small samples of participants recruited from a single source, raising serious concerns about the generalizability of these results in clinical practice. Thus, it has become critical to re-evaluate the efficacy of various common EEG features for MDD detection across large and diverse datasets. To address this issue, we collected resting-state EEG data from 400 participants across four medical centers and tested classification performance of four common EEG features: band power (BP), coherence, Higuchi’s fractal dimension, and Katz’s fractal dimension. Then, a sequential backward selection (SBS) method was used to determine the optimal subset. To overcome the large data variability due to an increased data size and multi-site EEG recordings, we introduced the conformal kernel (CK) transformation to further improve the MDD as compared with the healthy control (HC) classification performance of support vector machine (SVM). The results show that (1) coherence features account for 98% of the optimal feature subset; (2) the CK-SVM outperforms other classifiers such as K-nearest neighbors (K-NN), linear discriminant analysis (LDA), and SVM; (3) the combination of the optimal feature subset and CK-SVM achieves a high five-fold cross-validation accuracy of 91.07% on the training set (140 MDD and 140 HC) and 84.16% on the independent test set (60 MDD and 60 HC). The current results suggest that the coherence-based connectivity is a more reliable feature for achieving high and generalizable MDD detection performance in real-life clinical practice.

show abstract

Section: Cross Validation Used In Previous Work: More Detailed Review and Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Resting-State EEG Signal for Major Depressive Disorder Detection: A Systematic Validation on a Large and Diverse Dataset

Huang²,

Huang³

et al. 2021

Biosensors

View full text Add to dashboard Cite

show abstract

“…Xu et al [18] created an intelligent human-computer interaction (HCI) system for depression recognition, in which an audio-depression regression model determines the prevalence rate of depression based on CNN and long short-term memory (LSTM) network, and put forward a differential regression features extraction algorithm for multiscale audios to describe the features of nonpersonalized audios. Akbari et al [19] proposed a novel depression detection approach based on reconstructed phase space (RPS) and geometric features of electroencephalogram (EEG) signals, plotted the RPS maps of the EEG signals of 22 normal people and 22 depression patients in two-dimensional (2D) space, and extracted 34 geometric features according to the map shapes. It was learned that, compared with the RPS of depression patients' EEG signals, the RPS of normal people's EEG signals has a high regularity, limited changes, and predictable shapes.…”

Section: Introductionmentioning

confidence: 99%

Elderly Depression Recognition Based on Facial Micro-Expression Extraction

Huang¹

2021

View full text Add to dashboard Cite

Depression leads to a high suicide rate and a high death rate. But the disease can be cured if recognized in time. At present, there are only a few low-precision methods for recognizing mental health or mental disorder. Therefore, this paper attempts to recognize elderly depression by extracting facial micro-expressions. Firstly, a micro-expression recognition model was constructed for elderly depression recognition. Then, a jump connection structure and a feature fusion module were introduced to VGG-16 model, realizing the extraction and classification of micro-expression features. After that, a quantitative evaluation approach was proposed for micro-expressions based on the features of action units, which improves the recognition accuracy of elderly depression expressions. Finally, the advanced features related to the dynamic change rate of depression micro-expressions were constructed, and subjected to empirical modal decomposition (EMD) and Hilbert analysis. The effectiveness of our algorithm was proved through experiments.

show abstract