A machine learning approach for detection of claustrophobic brain activity in electroencephalography

Comprehensive Summary

This study, conducted by Rezaei et al., analyzes the abilities of machine learning models in classifying claustrophobic brain activity using electroencephalogram (EEG) data. From the a sample of participants with self-identified claustrophobia (n = 9) and healthy controls (n = 13), EEG readings were collected and pre-processed and organized into 5 types of EEG signals: delta waves, theta waves, alpha waves, beta waves, and gamma waves. Afterwards, the data was used to analyze classical machine learning models (k-Nearest Neighbors (kNN), Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF)) and deep learning models (Multi-layer Perceptron (MLP), Hybrid CNN-BiLSTM Deep Classifier). From the results, it was found that between control and claustrophobic groups, there were significant differences in EEG signal patterns within different regions of the brain: within the frontal lobe, the claustrophobic group experienced higher levels of delta waves than the control, and within the occipital lobe, the control group experienced higher alpha waves than the claustrophobic group. For the machine learning analysis, it was found that among all EEG signal types, MLP and CNN-BiLSTM models consistently performed well at accuracies >90%. DT performed significantly worse than other models in classifying claustrophobic brain activity with delta and theta signals with an accuracies between 75 - 80%. Overall, the models were able to classify claustrophobic brain activity at 70 - 95% accuracy.

Outcomes and Implications

Phobias are classified as extreme, disruptive fears of different things or conditions, such as animals, activities, and perceived free space. From recent studies, it was found that subjects with phobias generally exhibited differences in EEG patterns compared to control subjects. However, there has been a lack of studies investigating the differential EEG patterns among claustrophobia. To address this shortcoming in the literature, Rezaei et al. aimed to evaluate both the EEG patterns among claustrophobic participants and the potential of machine learning in assisting with the detection of claustrophobia.