- Proposed transformer using Maximally Collapsing Metric Learning applied to EEG feature matrices and compared with optimized Tunable Q-Wavelet Transform for schizophrenia identification.
- Eight nonlinear features were extracted from both MCML and TQWT approaches and classified using k-NN, Optimized Decision Tree and Deep Neural Network.
- Exceptional performance achieved on a single electrode channel with accuracies of 100%, 99.94% and 99.91%, demonstrating high discriminative power.
Comput Methods Biomech Biomed Engin. 2026 Oct 9:1-17. doi: 10.1080/10255842.2026.2743643. Online ahead of print.
ABSTRACT
Electroencephalography (EEG) is a noninvasive technique for recording brain electrical activity and supporting the assessment of psychiatric disorders. This study proposes a framework for distinguishing individuals with Schizophrenia (SZ) from healthy subjects using EEG data. The approach employs a transformer based on Maximally Collapsing Metric Learning (MCML) applied to feature-vector matrices and compares it with optimized Tunable Q-Wavelet Transform (TQWT) decomposition. Eight nonlinear features are extracted from both approaches and evaluated using *k*-Nearest Neighbor (*k*-NN), Optimized Decision Tree (ODT), and Deep Neural Network (DNN) classifiers. Results achieve accuracies of 100%, 99.94%, and 99.91%, respectively, using one electrode channel.
PMID:42854099 | DOI:10.1080/10255842.2026.2743643
Share Evidence Blueprint
Save to Google Notes
Search Google Scholar
Save as PDF
⭐ My Revision List
