Abstract:Early detection and accurate diagnosis of mental disorders is difficult due to the complexity of the diagnostic process, resulting in conditions being left undiagnosed or misdiagnosed. Previous studies have demonstrated that features of Electroencephalogram (EEG) data, such as Power Spectral Density (PSD), are useful biomarkers for indicating the onset of various mental disorders. Existing models using EEG data are typically trained to distinguish between healthy and afflicted individuals, and they are unable to distinguish between individuals with different disorders. We propose MENTAL (Multi-modal EEG NEO-FFI with Trained Attention Layer) to predict an individual's mental state using both EEG and Neo-Five Factor Inventory (NEO-FFI) personality data. We include an attention layer that captures the interactions between personality traits and PSD features, and emphasizes the important PSD features. MENTAL features a Recurrent Neural Network (RNN) to model the temporal nature of EEG data. We train our model with the Two Decades Brainclinics Research Archive for Insights in Neuroscience (TDBRAIN) dataset, which consists of 1274 healthy and psychiatric individuals including over 30 different diagnoses. MENTAL is able to achieve 93.3% accuracy when trained to classify between healthy individuals and those with ADHD. When trained to identify individuals with ADHD from among 33 disorder classes, MENTAL improves accuracy from 70.5 to 81.7%. MENTAL also demonstrates over 20% improvement in predictive accuracy when trained for MDD prediction. For the multi-class classification task of three classes, MENTAL improves accuracy by 4%, and for five classes, by nearly 8%. MENTAL is the first multi-modal model that utilizes both EEG and NEO-FFI data for the task of mental disorder prediction. We are one of the first groups to utilize TDBRAIN for automated disorder classification. MENTAL is accessible and cost-effective, as EEG is more affordable than other neuroimaging methods such as MRI, and the NEO-FFI is a self- reported survey. Our model can lead to acceptance and support for individuals living with mental health challenges and improve quality of life in our society.

Transformers for autonomous recognition of psychiatric dysfunction via raw and imbalanced EEG signals

DISTINGUISHING BIPOLAR DEPRESSION FROM MAJOR DEPRESSIVE DISORDER USING FNIRS AND DEEP NEURAL NETWORK

Major Depressive Disorder Detection Using Effective Connectivity of EEG Signals and Deep Learning Transformer Model

An Extended Computer-Aided Diagnosis System for Multidomain EEG Classification

Automated Detection of Major Depressive Disorder With EEG Signals: A Time Series Classification Using Deep Learning

Depression Diagnosis and Drug Response Prediction via Recurrent Neural Networks and Transformers Utilizing EEG Signals

Psychiatric disorders from EEG signals through deep learning models

Classification of Attention Deficit/Hyperactivity Disorder based on EEG signals using a EEG-Transformer model

Efficacy of Transformer Networks for Classification of Raw EEG Data

An attention-based deep learning approach for the classification of subjective cognitive decline and mild cognitive impairment using resting-state EEG

Exploring new horizons in neuroscience disease detection through innovative visual signal analysis

A machine learning based depression screening framework using temporal domain features of the electroencephalography signals

Deep Depression Detection with Resting-State and Cognitive-Task EEG

Automated Rest EEG-Based Diagnosis of Depression and Schizophrenia Using a Deep Convolutional Neural Network

Automatic Detection of Schizophrenia Based on Spatial–temporal Feature Mapping and LeViT with EEG Signals

Automatic Classification of EEG Signals Via Deep Learning.

AI-Enhanced Mental Health Diagnosis: Leveraging Transformers for Early Detection of Depression Tendency in Textual Data

Multi-modal EEG NEO-FFI with Trained Attention Layer (MENTAL) for mental disorder prediction

EEGformer: A transformer-based brain activity classification method using EEG signal

Automatic Analysis of EEGs Using Big Data and Hybrid Deep Learning Architectures

Diagnostic deep learning algorithms that use resting EEG to distinguish major depressive disorder, bipolar disorder, and schizophrenia from each other and from healthy volunteers