[Paper Review] Power Spectral Density-Based Resting-State EEG Classification of First-Episode Psychosis

Sadi Md. Redwan, Md Palash Uddin+2 more|arXiv (Cornell University)|Nov 23, 2022

EEG and Brain-Computer Interfaces4 citations

TL;DR

This study proposes a power spectral density (PSD)-based classification framework using resting-state EEG to distinguish first-episode psychosis (FEP) patients from healthy controls. Employing machine learning models—including Gaussian Process Classifier (GPC)—the approach achieves 95.78% specificity, demonstrating PSD as a robust biomarker for psychiatric disorder classification without stimulus dependency.

ABSTRACT

Historically, the analysis of stimulus-dependent time-frequency patterns has been the cornerstone of most electroencephalography (EEG) studies. The abnormal oscillations in high-frequency waves associated with psychotic disorders during sensory and cognitive tasks have been studied many times. However, any significant dissimilarity in the resting-state low-frequency bands is yet to be established. Spectral analysis of the alpha and delta band waves shows the effectiveness of stimulus-independent EEG in identifying the abnormal activity patterns of pathological brains. A generalized model incorporating multiple frequency bands should be more efficient in associating potential EEG biomarkers with First-Episode Psychosis (FEP), leading to an accurate diagnosis. We explore multiple machine-learning methods, including random-forest, support vector machine, and Gaussian Process Classifier (GPC), to demonstrate the practicality of resting-state Power Spectral Density (PSD) to distinguish patients of FEP from healthy controls. A comprehensive discussion of our preprocessing methods for PSD analysis and a detailed comparison of different models are included in this paper. The GPC model outperforms the other models with a specificity of 95.78% to show that PSD can be used as an effective feature extraction technique for analyzing and classifying resting-state EEG signals of psychiatric disorders.

Motivation & Objective

• To investigate whether resting-state EEG power spectral density (PSD) features can effectively differentiate first-episode psychosis (FEP) patients from healthy controls.
• To evaluate the performance of multiple machine learning models—random forest, support vector machine, and Gaussian Process Classifier (GPC)—in classifying FEP using PSD features.
• To establish a comprehensive preprocessing pipeline for PSD analysis in resting-state EEG data to ensure reliable feature extraction.
• To identify the most effective frequency bands (e.g., alpha, delta) for distinguishing pathological brain activity in FEP using stimulus-independent EEG signals.

Proposed method

• Acquisition of resting-state EEG data from FEP patients and healthy controls without task-based stimulation.
• Application of spectral analysis to compute power spectral density (PSD) across multiple frequency bands (e.g., delta, alpha).
• Preprocessing steps including artifact removal, filtering, and segmentation to ensure signal quality before PSD estimation.
• Feature extraction using PSD values across standardized frequency bands as input to machine learning models.
• Training and validation of three classifiers: random forest, support vector machine, and Gaussian Process Classifier (GPC).
• Model evaluation using metrics such as specificity, sensitivity, and area under the ROC curve to determine optimal classification performance.

Experimental results

Research questions

• RQ1 Can power spectral density (PSD) features from resting-state EEG reliably distinguish first-episode psychosis (FEP) patients from healthy controls?
• RQ2 Which machine learning model—random forest, support vector machine, or Gaussian Process Classifier (GPC)—yields the highest classification accuracy for FEP using PSD features?
• RQ3 Which frequency bands (e.g., delta, alpha) contribute most significantly to the discrimination of FEP from healthy controls in resting-state EEG?
• RQ4 How effective is a generalized multi-band PSD approach in identifying potential EEG biomarkers for FEP compared to single-band analysis?
• RQ5 What preprocessing pipeline optimizes PSD estimation and classification performance in resting-state EEG for psychiatric disorder detection?

Key findings

• The Gaussian Process Classifier (GPC) model achieved the highest specificity of 95.78% in distinguishing first-episode psychosis (FEP) patients from healthy controls.
• Spectral analysis of alpha and delta band power showed significant differences in resting-state EEG between FEP patients and healthy individuals.
• The use of multiple frequency bands in a generalized PSD-based model improved the identification of pathological brain activity patterns in FEP.
• Resting-state EEG signals, when analyzed via PSD, demonstrated sufficient discriminative power for psychiatric disorder classification without requiring task-based stimuli.
• The comprehensive preprocessing pipeline enhanced signal quality and contributed to the robustness of PSD feature extraction and model performance.
• The study confirms that PSD-based features from resting-state EEG are a viable and effective method for identifying potential biomarkers of FEP.