[1L5-OS-18b] 生体信号を活用した医療・ヘルスケアAI

[1L5-OS-18b-04] Construction of a High Accuracy Depression Detection Model using Single Electrode EEG Data with Machine Learning

〇Kei Suzuki¹, Midori Sugaya¹ (1. Shibaura Institute of Technology)

Keywords:AI, EEG, Depression

Recently, study of constructing a depression detection model applying machine learning using EEG signals from a small number of electrodes has been conducted. However, the accuracy is about 80%, and a highly accurate method is not yet obvious. In this study, we propose a method to construct a depression detection model with high accuracy by applying machine learning using single electrode EEG signals. To clean the EEG signals, we apply filtering, epoch rejection, and outlier detection. Then, we calculated the complexity of the EEG signals as EEG indices. The EEG indices were trained by LightGBM, a decision tree-based model. Cross-validation results showed that the accuracy of 90% was achieved in the binary classification of depressed patients and healthy controls. These results suggest that the proposed method is effective in detecting depression.