Enhanced Performance by Interpretable Low-Frequency Electroencephalogram Oscillations in the Machine Learning-Based Diagnosis of Post-traumatic Stress Disorder

Machine learning-based differentiation of schizophrenia and bipolar disorder using multiscale fuzzy entropy and relative power from resting-state EEG

Schizophrenia (SZ) and bipolar disorder (BD) pose diagnostic challenges due to overlapping clinical symptoms and genetic factors, often resulting in misdiagnosis and suboptimal treatment outcomes. This study aimed to identify EEG-based biomarkers that can differentiate SZ from BD using multiscale fuzzy entropy (MFE) and relative power (RP) analyses and to evaluate their diagnostic utility using machine learning. EEG data were collected from 65 patients with SZ and 49 patients with BD under resting-state conditions. The MFE and RP were calculated for the bilateral frontal, central, and parietal regions using 30 s EEG segments. For MFE, the band-scale fuzzy entropy (FuzzyEn) was determined across the theta, alpha, beta, and gamma bands based on simulation results demonstrating an inverse relationship between scale factors and frequency components. RP was derived by segmenting the EEG data into 2 s intervals with a 500 ms moving window. A support vector machine (SVM) was used to differentiate between patients with SZ and BD based on band-scale FuzzyEn and RP. The SVM classifier achieved an accuracy of 78.94%, a sensitivity of 81.53%, and a specificity of 75.51%. Patients with SZ showed higher theta-scale FuzzyEn in the right frontal, left central, and bilateral parietal regions; higher alpha-scale FuzzyEn in the right parietal region; and increased theta power in the bilateral frontal, central, and right parietal regions. These differences remained robust after controlling for medication effects. These findings demonstrate the potential of combining MFE, RP, and machine learning to differentiate between SZ and BD, contributing to improved diagnostic precision in psychiatric disorders.

Systematic review of machine learning in PTSD studies for automated diagnosis evaluation

Post-traumatic stress disorder (PTSD) is frequently underdiagnosed due to its clinical and biological heterogeneity. Worldwide, many people face barriers to accessing accurate and timely diagnoses. Machine learning (ML) techniques have been utilized for early assessments and outcome prediction to address these challenges. This paper aims to conduct a systematic review to investigate if ML is a promising approach for PTSD diagnosis. In this review, statistical methods were employed to synthesize the outcomes of the included research and provide guidance on critical considerations for ML task implementation. These included (a) selection of the most appropriate ML model for the available dataset, (b) identification of optimal ML features based on the chosen diagnostic method, (c) determination of appropriate sample size based on the distribution of the data, and (d) implementation of suitable validation tools to assess the performance of the selected ML models. We screened 3186 studies and included 41 articles based on eligibility criteria in the final synthesis. Here we report that the analysis of the included studies highlights the potential of artificial intelligence (AI) in PTSD diagnosis. However, implementing AI-based diagnostic systems in real clinical settings requires addressing several limitations, including appropriate regulation, ethical considerations, and protection of patient privacy.