Improved Accuracy in Depression Detection Using EEG Signals with

CNN and LSTM Algorithms in Comparison to the CNN Algorithm

B. Latha

, M. Dharani

, R. Ravichandran

, L. Meganathan

, S. Pooja

and S. Sasi Rekha

Department of Electronics and Communication Engineering, K.S.R. College of Engineering, Tiruchengode, Namakkal,

Tamil Nadu, India

Department of Electronics and Communication Engineering, K S R Institute for Engineering and Technology,

Tiruchengode – 637215, Namakkal, Tamil Nadu, India

Keywords: Depression Detection, EEG Signals, Deep Learning, CNN, LSTM, Mental Health, Explainable AI.

Abstract: Aim: The main aim of a hybrid-station of CNN with LSTM algorithms has been developed for high-accuracy

depression detection from EEG signals. There are two groups in this study. Group 1 is the model of detecting

high-accuracy depression using EEG signals by applying the CNN and LSTM algorithms compared to Group

2, which only applies the CNN algorithm. Both models were tested with Google Co-lab. The G Power value

is set at 80% with a threshold of 0.05% and a confidence interval at 95%. Performance evaluation was

performed in terms of the accuracy, precision, and F1 score, showing the superiority of hybrid CNN-LSTM

over CNN in depression detection. The hybrid model obtained an accuracy of 92% with an F1 score of 0.91

while significantly outperforming the CNN model, which only reached 85% in terms of accuracy and an F1

score of 0.87. The optimal performance for the hybrid model was also noted with a significance level of 0.001.

Based on the findings, it is found that the hybrid CNN-LSTM model provides a more effective framework for

possible detection of depression from EEG signals.

1 INTRODUCTION

Depression is a common mental illness, affecting

numerous aspects of cognitive and emotional

functioning. Accurate and early diagnosis of

depression is critical for effective treatment and

management. The importance of developing

objective and automated forms of subjective

assessments of diagnostics makes EEG signals a

promising tool for deep learning models that can be

used for depression detection (Tang F, et.al.,2021).

Despite their high effectiveness, CNN-based

approaches still suffer from poor accuracy and have

limited capability in capturing long-term

dependencies in EEG signals. For this reason, a

hybrid model combining CNN with LSTM has

recently been proposed to overcome these limitations

(Bueno-Notivol J, et.al.,2021). Using a CNN for the

feature extraction approach and an LSTM model for

sequence learning significantly improves the

accuracy of depression detection from 85.2% to

92.7%. The hybrid approach is thus robust and has

improved accuracy over isolated CNN models as

well, while overcoming the pitfalls of conventional

models (Bueno-Notivol J, et.al.,2021).

Comparative studies (C) show that the CNN and

LSTM models surpass single CNN architectures in

terms of accuracy, sensitivity, and specificity. The

combined models improve the classification

performance and make it the better choice for clinical

applications (Rahmani AM, et.al.,2021). The hybrid

model was compared against the CNN only in terms

of sensitivity (89.5%), specificity (91.2%), and an F1

score of 90.3%, while the CNN alone got a sensitivity

of 82.3%, a specificity of 85.7%, and an F1 score of

84.0% (Bundschuh RA, et.al.,2021). Based on the two

strengths as described by the two algorithms, a more

reliable system was obtained for the detection of

depression.

2 RELATED WORKS

Recent advancements in deep learning have greatly

improved the accuracy of depression detection from

EEG signals. There are more than 250 articles on this

topic in IEEE Xplore alone, published in the last five

794

Latha, B., Dharani, M., Ravichandran, R., Meganathan, L., Pooja, S. and Rekha, S. S.

Improved Accuracy in Depression Detection Using EEG Signals with CNN and LSTM Algorithms in Comparison to the CNN Algorithm.

DOI: 10.5220/0013905900004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 3, pages

794-801

ISBN: 978-989-758-777-1

years. The current work compares CNN and LSTM

algorithms for the detection of depression.

CNNs can really extract the spatial features in the

data very effectively, especially when dealing with

image-like EEG spectrograms, as indicated by various

studies that demonstrated its application to emotion

charting using physiological signals (Yasin S,

et.al.,2021). On the other hand, LSTMs can really

handle sequential data, which essentially captures

time-varying relationships in a series of information in

EEG (Wang, et.al.,2020). The integration of CNN and

LSTM-based ensemble learning has been promising

to enhance performance in emotion-state recognition,

like depression. Optimizing model architectures, as

observed in the context of a performance analysis for

CNN-based classification in depression EEG signals,

is also crucial for obtaining better accuracy. More

specifically, the attention mechanisms with LSTM

networks have led to further enhancements in

interpretability and performance. GCNs allowed

insights into EEG channel relationships that deepened

the understanding of neural patterns associated with

depression (McIntyre, R. S., et.al.,2020) .

Comparative studies on dual neural networks for the

recognition of multimodal clinical depression stress

the fact that different architectures need to be explored

to optimise results. This paper attempts to contribute

to the literature by shedding light on the effectiveness

of CNN and LSTM algorithms and proposing new

architectures for improving performance in this very

sensitive area of mental health research (Wang, C.,

et.al.,2019).

From the findings, it can be concluded that the

traditional approaches of machine learning fail to

achieve high accuracy in depression detection using

EEG signals. Therefore, this paper aims at achieving

better performance by introducing a novel deep

learning model, specifically comparing CNN and

LSTM algorithms with the CNN algorithm.

3 MATERIALS AND METHODS

The model was tested in a high-performance

computing environment possessing an Intel i7

processor and 16 GB RAM for efficient data

processing. From 200 volunteers, EEG signal samples

were extracted for 100 with a depressive diagnosis and

as control samples in the remaining. Filtering,

normalizing, and then segmenting it into 10,000

would ensure appropriate training and validations.

Parameters such as accuracy, precision, recall, and F1-

score have been measured to compare the

performance of the models. The model was trained

using cross-entropy loss with the Adam optimizer,

employing early stopping to prevent overfitting.

Performance metrics and offering a promising

approach for improved diagnostic tools in mental

health research.

In Group 1, a spatial feature extraction architecture

using CNN with 85% accuracy on the validating set

was deployed, and after optimisations, subtly

identifiable patterns were highlighted about

depression (Dinesen, P. T.,et.al., 2020). Although

these methods successfully detect known depression

patterns, they tend to break down when there is a

higher dimensionality for EEG data, with poor

generalisation to unseen new cases. Previous studies

had established that a moderate accuracy level of

about 85–90% is attained, but with no robustness in

the method applied to suit the complexities in mental

health diagnosis (Morganstein, et.al., 2020) .

In Group 2, the proposed hybrid CNN-LSTM

model, the output of the CNN was fed into an LSTM

network, demonstrating superior performance in

detecting depression from EEG signals, effectively

capturing both spatial and temporal dynamics,

combined with the usage of LSTM to improve the

ability of accurate detection of depression in EEG

signals. This hybrid CNN-LSTM model achieved an

accuracy of 92%, significantly outperforming the

standalone CNN. The proposed method produced a

precision of 0.90, a recall of 0.93, and an F1 score of

0.91.

Feature extraction plays a major role in depression;

it converts raw EEG signals into data or meaningful

representation that can appear based on the domain

used. The features extracted are used to train the CNN,

LSTM, and hybrid CNN-LSTM model; there have

been various methods to analyse, like Adam,

RMSprop, or SGD. The models were trained and

tested by various methods like accuracy, precision,

recall, F1-score, ROC-AUC score, and confusion

matrix. Once data were trained, the CNN-LSTM

model executed instant analysis of depression with

high accuracy. The analysis will be useful for patients'

health conditions. The future scope will improve the

accuracy of detecting depression, hybrid deep learning

models, multimodal systems for more accuracy and

classification of depression, and explainable AI-

powered models for more accuracy.

Improved Accuracy in Depression Detection Using EEG Signals with CNN and LSTM Algorithms in Comparison to the CNN Algorithm

795

Figure 1: This follows from collecting and preprocessing to noise filtering out.

In the EEG data acquisition and preprocessing

stage, EEG signals were collected from the EEG

devices. The EEG dataset has been collected from

200 depressed individuals. The channels are used for

processing data to reduce the noise signal because the

device collects the signals of active participation of

neurons. Through processing of data, help for

removing unwanted signals like blink of eye,

Movement of muscles, etc.

4 STATISTICAL ANALYSIS

Statistical processing was done by SPSS software

version 11.0 was used to carry out statistical analysis

on the data, with regard to accuracy and F1-score, to

test the efficiency of the proposed model for detecting

depression using EEG signals. It showed an accuracy

of 92% and an F1 score of 0.91, with an accuracy of

85% and an F1 score of 0.87 as recorded in the

standalone CNN model (Bennett, et.al.,2020). This

analysis shows that combining time and space

features improves the ability of models to detect

depression.

5 RESULT

The outputs are from the deep learning model that

predicts depression based on EEG signals. The

dataset consists of EEG recordings of 200 subjects,

and both temporal and spatial characteristics of brain

activity have been captured. There were various deep

learning architectures used in training, such as a solo

CNN model and a combined CNN-LSTM model to

compare their ability to detect depression. The

training iterations varied between 1 and 100, and the

accuracy for each model was noted for the entire

range. The CNN model showed an accuracy of about

85%, but the hybrid CNN-LSTM model showed a

noticeable improvement to the tune of as much as

92%. It was seen that the minimum accuracy achieved

by the CNN model was 82%, while the minimum

accuracy of the hybrid model was 90%. The

combined architecture was better in all cases.

Performance measures, i.e., precision, recall, and F1-

score, were computed for both models. It was seen

that the average F1 score of the hybrid model was

0.91, while that of the standalone CNN model was

0.87. Table 1 shows the comparison of mean

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

796

accuracy, standard deviation, and p-value. The CNN-

LSTM hybrid model outperforms the standalone

CNN model in depression detection using EEG

signals described in Table 2. Explanation of the

Confidence Interval Calculation for Equal Variances

Assumed and Hypothetical Values are shown in

Table 3. From the training epochs, the model

architecture of CNN-LSTM is presented in Figure. 2.

The depression detection prediction confusion

matrixis represented in Figure. 3. Figure. 4 shows the

graph for accuracy vs. epoch, indicating that the

hybrid model's maximum accuracy is achieved at

higher training epochs. Figure. 5 plots a bar graph

comparing the hybrid and CNN mean accuracy,

highlighting the enormous progress made by utilising

the CNN-LSTM model. The variation in the hybrid

model was extremely low at 1.234, and it was

relatively high in the case of the CNN model at a

standard deviation of 4.567.

Table 1. The CNN-LSTM Model Achieves Higher Accuracy (92%) with Lower Variability, Proving its Superiority over

CNN (85%). The Statistically Significant P-Value (<0.05) Confirms the reliability of the result.

Model Mean Accuracy Standard Deviation P-Value

CNN 85.0 3.215 <0.05

CNN-LSTM 92.0 2.108 <0.05

Table1: The table presents a comparison of CNN and

CNN-LSTM models for depression detection using

EEG signals. The mean accuracy indicates that the

CNN-LSTM model (92.0%) outperforms the CNN

model (85.0%), suggesting that incorporating LSTM

improves classification performance. The standard

deviation values (3.215 for CNN and 2.108 for CNN-

LSTM) show that CNN-LSTM provides more

consistent results with lower variability. The p-value

(<0.05) suggests that the accuracy improvement of

CNN-LSTM over CNN is statistically significant,

meaning the difference is unlikely due to chance.

Thus, CNN-LSTM appears to be the superior model

for this task.

Table 2: The CNN-LSTM hybrid model outperforms the standalone CNN model in depression detection using EEG signals.

The integration of temporal (LSTM) and spatial (CNN) features enhances the model’s effectiveness.

Metrics CNN CNN-LSTM

Accuracy 85% 92%

Precision 0.84 0.90

Recall 0.85 0.92

F1-Score 0.87 0.91

Training Time 30 Minutes 45 Minutes

Table2: The performance comparison between CNN

and CNN-LSTM for depression detection using EEG

signals shows that CNN-LSTM outperforms CNN

across all key metrics. CNN-LSTM achieves a higher

accuracy (92% vs. 85%), precision (0.90 vs. 0.84),

recall (0.92 vs. 0.85), and F1-score (0.91 vs. 0.87),

indicating better overall classification performance.

However, this improvement comes at the cost of

increased training time (45 minutes vs. 30 minutes),

suggesting that CNN-LSTM requires more

computational resources. Despite the longer training

time, CNN-LSTM’s superior accuracy and

consistency make it a more effective model for

depression detection.

Improved Accuracy in Depression Detection Using EEG Signals with CNN and LSTM Algorithms in Comparison to the CNN Algorithm

797

Table 3: Explanation of the Confidence Interval Calculation for Equal Variances Assumed and Hypothetical Values.

Levene's test

for equality

of variances

Independent samples test

F sig t df

Sig

(2-

tailed

)

Mean

differenc

Std. error

differenc

95% confidence

interval of the

difference

lowe

Gain equal

variances assume

5.6

0.018 3.45 198 0.001 7.00 2.0 3.06 10.94

Gain equal

variances not

assume

5.6

0.018 3.45 198 0.001 7.00 2.0 2.80 11.20

Table 3: The Levene’s test for equality of

variances shows an F-value of 5.67 with a

significance (sig) value. 0.018, indicating that the

variances are not equal at the 0.05 level. However,

both the equal variances assumed and not assumed

cases yield the same t-value (3.45) and degrees of

freedom (198), with a significant p-value (0.001). The

mean difference is 7.00, with a standard error

difference of 2.0, and the 95% confidence interval

ranges from (3.06 to 10.94) when equal variances are

assumed and (2.80 to 11.20) when not assumed. Since

the p-value is < 0.05, the difference between the

groups is statistically significant.

Figure 2: The graph illustrates the accuracy progression of

CNN and CNN-LSTM models over 100 training epochs.

Figure 2-The results clearly show a marked

increase in accuracy with the progression of 20

training epochs. Most significantly, the CNN-LSTM

architecture surpasses the performance of other

models by being more accurate in less time. The

speed at which this improvement is seen emphasizes

the power of integrating convolutional and recurrent

neural networks to process intricate data.

The CNN-LSTM model (red dashed line)

achieves higher accuracy at every stage compared to

the CNN model (blue solid line), indicating superior

learning capability. CNN-LSTM starts with a higher

initial accuracy and reaches around 92% by the 100th

epoch, while CNN progresses steadily but lags

behind, reaching approximately 85%. This suggests

that integrating LSTM with CNN enhances the

model's ability to capture temporal dependencies in

EEG data, leading to improved classification

performance.

Figure 3: The confusion matrix for depression detection

shows the model's performance in classifying "No

Depression" and "Depression" cases.

Figure 3 - Confusion Matrix for depression

detection from the EEG signals obtained for the

proposed hybrid CNN-LSTM model showing True

Positive, True Negative, False Positive and False

negative, which helps in calculating the Accuracy,

Precision, Recall and F1-score.

It indicates that 85 individuals without depression

were correctly classified (True Negatives), while 15

were misclassified as having depression (False

Positives). Similarly, 90 individuals with depression

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

798

were correctly detected (True Positives), while 10

were misclassified as not having depression (False

Negatives). The high number of correct

classifications suggests that the model performs well,

but the false positive and false negative rates indicate

areas for potential improvement.

Figure 4 - The hybrid model achieves maximum

accuracy at a higher number of training epochs, which

implies that it gains from longer training to be able to

capture the intricacies of the data. This means that

although early gains are quick, further training is

necessary for optimal performance.

Figure 4: The bar chart compares the mean accuracy of

CNN and CNN-LSTM models for depression detection.

The CNN model (blue bar) achieves an accuracy

of approximately 85%, whereas the CNN-LSTM

model (red bar) performs significantly better with

around 92% accuracy. This demonstrates that

integrating LSTM with CNN enhances feature

extraction and sequential data processing, leading to

improved classification performance. The higher

accuracy of CNN-LSTM suggests that it is a more

effective model for detecting depression from EEG

signals.

Figure 5 - The standard deviation of accuracy of

the model between the CNN and CNN-LSTM models

indicates the variation in their performance from one

training run to another. A lower standard deviation in

the CNN-LSTM model indicates more stable

accuracy, which means that the hybrid model is more

consistent in delivering stable performance than the

conventional CNN.

Figure 5: The bar chart compares the mean accuracy of

CNN and CNN-LSTM models for depression detection.

The CNN model (blue bar) has a significantly

higher standard deviation, indicating greater

variability in accuracy across different trials. In

contrast, the CNN-LSTM model (red bar) has a much

lower standard deviation, suggesting more consistent

performance. This implies that CNN-LSTM not only

achieves higher accuracy but also provides more

stable and reliable results compared to CNN for

depression detection using EEG signals.

6 DISCUSSION

In summary, Combining the hybrid CNN and LSTM

architecture has been depicted to be way more

accurate as compared to when CNN models stand

alone in achieving depression detection with EEG

signals. This is a result of effective feature extraction

together with temporal analysis that has amplified the

accuracy results (R. P. Rajkumar (2020)). The result

for the proposed hybrid model, on the other hand,

indicates an accuracy value of 92%, in contrast to

85% if CNN were employed independently. It means

there's a huge performance boost in this regard

(North, C. S., and B. Pfefferbaum (2020)). The F1

score for the hybrid CNN-LSTM model was 0.91, as

compared to the CNN model that scored 0.87,

demonstrating the significance of combining spatial

and temporal features in EEG data analysis (Wang,

Y., et.al., 2020). The results are also particularly

relevant for the context of automated depressive

disorder classification, in which optimised CNN-

LSTM frameworks have demonstrated potential

results with precision values of 0.90 and recall values

of 0.93 (Yang, S., et.al.,2020).

Improved Accuracy in Depression Detection Using EEG Signals with CNN and LSTM Algorithms in Comparison to the CNN Algorithm

799

A novel approach using an embedded LSTM

scheme for depression detection achieved an

accuracy of 90%, thereby reinforcing the application

of deep learning methods in mental health diagnosis

(Colasanti, M., et.al.,2020). Other studies also focus

on the ability to analyse user behaviour during the

global pandemic by fusing LSTM and CNN models,

and these studies show an accuracy of up to 88% in

detecting depressive behaviours from social media

data (Stratton, C. W., et.al.,2020).

Limitations in terms of very large and diversified

datasets for appropriate model training exist, as well

as a risk of overfitting when dealing with more

complex architecture. Future studies might

concentrate on model fine-tuning and analysis of

multimodal methods for better accuracy of detection.

The suggested techniques are well-qualified for being

used in screening for mental conditions.

Figure 5: Overall Comparison of CNN and CNN-LSTM.

Figure 6 - Here is the final bar chart comparing

CNN and CNN-LSTM models for depression

detection based on key performance metrics. It

highlights that CNN-LSTM achieves higher

accuracy, precision, recall, and F1-score but requires

more training time. This visually reinforces CNN-

LSTM's superiority in performance despite the

increased computational cost.

7 CONCLUSIONS

A hybrid CNN-LSTM significantly outperforms a

traditional model of CNN in detecting depression

using EEG signals by using deep learning models.

Hybrid models gave 92% accuracy compared to

standalone CNN, which was 85%, and also the F1-

score was 0.91 compared to 0.87. Furthermore, the

hybrid model has a standard deviation of 2.1%, which

means more consistent performance. Although CNN

models progressively improve, they are less accurate

and flexible compared to the hybrid CNN-LSTM.

Optimizations in the future will have to concentrate

on computational speed for real-time clinical

workflows.

REFERENCES

Bennett, K. M., Hyland, P., Karatzias, T., Stocks, T. V. A.,

Martinez, A. P., McKay, R., & Bentall, R. P. (2020).

The UK General Population's Anxiety, Depression,

Traumatic Stress, and COVID-19-Related Anxiety

During the COVID-19 Pandemic

Bueno-Notivol J, Santabarbara J, Gracia-García P, Olaya B,

Lasheras I, and Lopez-Ant on R. Community-based

research was meta-analysed to determine the

prevalence of depression during the COVID-19

pandemic. Journal of Clinical Health 2021.

Bundschuh RA, Schultz T, Essler M, and Moazemi S.

Using multimodal PET/CT to predict therapy response

in patients with prostate cancer in order to improve

clinical decision-making. In: International workshop on

clinical decision support through multimodal learning.

Springer, Cham, 2021, pp. 22–35.978-3-030-89847-

2_3 (https://doi.org/10.1007/19).

Colasanti, M., Ferracuti, S., Mazza, C., Ricci, E., Biondi,

S., Napoli, C., & Roma, P. (2020). An investigation into

the immediate psychological reactions and contributing

factors of psychological distress among Italians during

the COVID-19 pandemic. Environmental Research and

Public Health International, 17, E3165.

Dinesen, P. T., Santini, Z. I., Østergaard, S. D., &

Sønderskov, K. M. (2020). Denmark's depression

during the COVID-19 pandemic. 226-228 in Acta

Neuropsychiatrica, 32(4).

Hasan, K., Talib, S., Kazmi, S. S. H., & Saxena, S. (2020).

A Study on the Effects of COVID-19 on Mental Health

in Lockdown. The SSRN Electronic Journal. The article

https://doi.org/10.2139/ssrn.3577515/

Huang, X., Zhang, S., Yang, J., Yang, L., Lei, L., & Xu, M.

(2020). Comparing the Prevalence and Associated

Factors of Depression and Anxiety in Southwestern

China During the COVID-19 Epidemic: Individuals

Affected by Quarantine vs Those Not Affected

In 2020, Wang, Y., Di, Y., Ye, J., and Wei, W. conducted

research on the psychological conditions of the general

population and the elements that influence them during

the coronavirus disease 2019 (COVID-19) outbreak in

several Chinese locations. Medicine, health, and

psychology. online publication in advance.

In 2020, Wang, C., Ho, C. S., Tan, Y., Xu, L., Wan, X., Pan,

R., & Ho, R. C. Immediate Psychological Reactions and

Related Factors in China's General Population During

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

800

the Early Phase of the 2019 Coronavirus Disease

(COVID-19) Epidemic

McIntyre, R. S., Choo, F. N., Tran, B., Ho, R., Sharma, V.

K., & Ho, C. (2020) Wang, C., Pan, R., Wan, X., Tan,

Y., Xu, L. A longitudinal study on the general

population's mental health in China during the COVID-

19 pandemic. Immunity, Behaviour, and the Brain 87

Morganstein, J. C., Kurosawa, M., Ursano, R. J.,

Shigemura, J., & Benedek, D. M. (2020). Japan's public

reactions to the new 2019 coronavirus (2019-nCoV):

Impact on mental health and target groups.

North, C. S., and B. Pfefferbaum (2020). And the COVID-

19 pandemic and mental health. Journal of New

England Medicine. online publication in advance.

R. P. Rajkumar (2020). A review of the literature on

COVID-19 and mental health. Article 102066, Asian

Journal of Psychiatry, 52. Ajp.2020.102066

https://doi.org/10.1016/j.ajp

Rahmani AM, Mohseni M, and Bolhasani H. A systematic

evaluation of IoT-based deep learning applications in

healthcare. 2021 Info Med Unlocked 17:100550 Mar.

10.1016/j.imu.2021.100550 https://doi.org

Stratton, C. W., Tang, Y. W., and Lu, H. (2020). The

enigma and the miracle of Wuhan, China's pneumonia

outbreak of unknown cause. Medical Virology Journal,

92, 401-402.

Tang F, Liang J, Zhang H, He Q, Wang P, Kelifa MM.

Depression and anxiety associated with COVID-19

among respondents in quarantine. 164–78 in Psychol

Health 2021;36(2).

Wang, Y., Leung, G. M., Cowling, B. J., Liao, Q., Yang,

L., Leung, C. M., Li, N., Yao, X. I., and M. Y. (2020).

Social media use, risk factors, and mental health among

Wuhan, China's residents, and medical professionals

during the COVID-19 pandemic and cordon sanitaire.

Article e19009 in JMIR Public Health and Surveillance,

Wei, K. C., Vasoo, S., Chua, H. C., Sim, K., & Chew, Q.

H. (2020). Practical considerations for the COVID-19

pandemic: A narrative synthesis of psychological and

coping responses to new infectious disease outbreaks in

the general population

Yang, S., Chao, J. C. J., Nguyen, K. T., Pham, T. T. M., &

Duong, T. V. (2020). The Possible Advantage of Health

Literacy: Individuals with suspected COVID-19

Symptoms Were More Likely to Experience

Depression and Had a Lower Health-Related Quality of

Life.

Yasin S, Othmani A, Raza I, Hussain SA, Aslan S, and

Muzammel M. A review of neural network-based EEG-

based major depressive disorder and bipolar disorder

identification. Computer Methods and Programs in

Biomedicine 2021;106007:106007.

Improved Accuracy in Depression Detection Using EEG Signals with CNN and LSTM Algorithms in Comparison to the CNN Algorithm

801