Depression assessment using integrated multi-featured EEG bands deep neural network models: Leveraging ensemble...

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Mental Status Assessment (MSA) is crucial in psychiatry.
• Electroencephalogram (EEG) is increasingly used for mental state and depression evaluation.
• Existing methods may lack accuracy or require complex setups.

Purpose of the Study:

• To develop a novel multi-tier ensemble learning approach for mental state and depression assessment using integrated EEG bands.
• To evaluate the efficacy of a Long Short-Term Memory (LSTM)-based Deep Neural Network (DNN) and Multiple Linear Regression (MLR) ensemble.
• To validate the method's performance using a simple, commercialized one-channel EEG sensor.

Main Methods:

• EEG signals were divided into eight sub-bands.
• LSTM-based DNN models were trained for each EEG band.
• A two-tier ensemble learning approach using MLR integrated multi-band models for assessment.

Main Results:

• The proposed method achieved high performance metrics: accuracy (0.897), F1-score (0.921), precision (0.935), NPV (0.829), recall (0.908), specificity (0.875), and AUC (0.8917).
• Data collected from 57 subjects (49 depressed, 18 healthy) using a single-channel EEG sensor at FP1.
• Demonstrated superior performance compared to other ensemble learning techniques.

Conclusions:

• The novel multi-tier ensemble learning approach effectively assesses mental state and depression levels.
• The method shows significant potential for enhancing depression assessment accuracy.
• Utilizing integrated EEG bands with ensemble learning offers a viable and accurate diagnostic tool.