A qeeg-based ml scoring model for stress levels: reliable data and relying on computational power
Poster No:
537
Submission Type:
Abstract Submission
Authors:
Gangyoung Lee1, Seungwan Kang2
Institutions:
1iMediSync, Seoul, Seoul, 2iMediSync Inc., Seoul, Korea, Republic of
First Author:
Co-Author:
Introduction:
In recent research, Electroencephalography (EEG) has been established as a powerful biomarker, already in use for predicting stress. However, previous studies often had limitations such as a small sample size , evaluations conducted exclusively on healthy individuals , or the classification of participants as healthy or stress-affected based on subjective evaluations using scales like the PSS-10 or PCL-5 by individuals or psychological experts. Such factors are well-known in previous research to potentially lead to random or ambiguous outcomes . Therefore, this study aims to develop a model capable of quantitatively evaluating stress levels based on the performance of a model algorithm, rather than relying on subjective assessments. By utilizing a sufficiently large dataset that satisfies representativeness and a reliable binary classification label determined by psychiatrists, the relatively low resolution of the label provides a simplified yet effective foundation for robust model development.
Methods:
A total of 355 participants were included, consisting of 179 healthy individuals and 176 individuals with stress-related conditions. The dataset was labeled as either healthy or stress-affected by experienced psychiatrists affiliated with Seoul National University Hospital and Boeun Hospital. Resting-state EEG data were collected for 5 minutes using a 19-channel 10-20 system. Preprocessing steps, including denoising, removal of bad epochs, independent component analysis (ICA), and the calculation of EEG metrics, were performed using iMediSync's iSyncBrain platform. The computed EEG metrics included absolute spectral band power (Delta, Theta, Alpha1, Alpha2, Beta1, Beta2, Beta3, Gamma) and ratios (TAR, TBR, DAR).
The stress level model was trained and compared using ensemble models predominantly utilized in EEG-based modeling, including RandomForest, GradientBoost, AdaBoost, CatBoost
Model performance was evaluated based on recall and F1-score, as the primary goal was to support disease screening and diagnosis. Additionally, stress levels were expressed as percentages (0 for healthy individuals, 100 for stress-affected individuals) and were used as the probabilities of classifying participants into the corresponding label. During testing, the classification probabilities for the two labels were examined to ensure that the binary classification at the 50-point threshold was achieved while maintaining a flexible distribution between 0 and 100 points, avoiding extreme outputs.
The stress level model was trained and compared using ensemble models predominantly utilized in EEG-based modeling, including RandomForest, GradientBoost, AdaBoost, CatBoost
Model performance was evaluated based on recall and F1-score, as the primary goal was to support disease screening and diagnosis. Additionally, stress levels were expressed as percentages (0 for healthy individuals, 100 for stress-affected individuals) and were used as the probabilities of classifying participants into the corresponding label. During testing, the classification probabilities for the two labels were examined to ensure that the binary classification at the 50-point threshold was achieved while maintaining a flexible distribution between 0 and 100 points, avoiding extreme outputs.
·Model Pipeline Overview
Results:
Among the 4 ensemble models, the RandomForest model demonstrated the best performance, with precision of 88%, recall of 86%, F1-score of 87%, and accuracy of 87%. When tested on the test data, the model was observed to exhibit a flexible distribution. The 10 features selected through FRE included Theta-T4, Theta-Pz, Alpha1-P4, Alpha2-O1, Alpha2-O2, Beta1-O2, TBR-F7, TBR-F8, TBR-Cz, and TBR-Pz. These results suggest that the right hemisphere's Theta and high-frequency bands, such as Alpha2 and above, predominantly influence the severity of stress.
·Model AOC curve (Test data)
Conclusions:
This study aimed to address limitations in previous research, such as small sample sizes, data composed solely of healthy individuals, and unclear stress level labeling. Based on the improved dataset, we developed a model that relies on computational methods rather than subjective evaluations to assess stress severity. The resulting model demonstrated sufficient performance and was found to exhibit a flexible distribution, avoiding extreme outcomes. Future research will focus on collecting more comprehensive data to cluster stress subtypes and evaluate the severity of each cluster.
Brain Stimulation:
Non-invasive Electrical/tDCS/tACS/tRNS
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 1
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 2
Keywords:
DISORDERS
Electroencephaolography (EEG)
Modeling
Other - qeeg, stress
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
Which processing packages did you use for your study?
Provide references using APA citation style.
Saeed, S. M. U., Anwar, S. M., Khalid, H., Majid, M., & Bagci, U. (2020). EEG based classification of long-term stress using psychological labeling. Sensors, 20(7), 1886.
Nielsen, L., & Kaszniak, A. W. (2007). Conceptual, theoretical, and methodological issues in inferring subjective emotion experience. Handbook of emotion elicitation and assessment, 361-375.