Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{205042,
        author = {Prateek Yadav and Dr. Divakar Singh and Dr. Amit Kr. Jha and Dr. Kamini Maheshwar},
        title = {Bi-Transnet: A Deep Learning Approach for RealTime Psychological Disorder Detection via Electroencephalogram},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {1},
        pages = {5191-5201},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=205042},
        abstract = {The integration of electroencephalogram (EEG) signal analysis, particularly motor imagery (MI)-based techniques, into brain-computer interfaces (BCI) targeting the frontal cortex has witnessed remarkable growth. Deep learning has emerged as a fundamental methodology for extracting features from EEG recordings. However, two primary challenges impede the application of these approaches for EEG signal acquisition: Firstly, establishing a robust learning model demands extensive annotated data, yet the majority of EEG signal datasets lack proper labeling, making manual categorization a formidable obstacle. Secondly, developing a comprehensive learning model from the ground up requires substantial computational resources and time investment.
To address these challenges, we present an architecturally enhanced framework designed to improve feature extraction and classification accuracy for real-time mental condition detection, with emphasis on EEG data analysis. This framework, termed the bi-directional transformer network (Bi-TransNet), surpasses current state-of-the-art methodologies by 4.82%, achieving an outstanding classification accuracy of 99.12% on the DEAP multichannel scalp sleep staging EEG dataset.
The model demonstrates exceptional performance metrics: a remarkably low false positive rate per hour (FPRh) of 0.42%, combined with sensitivity of 99.10% and specificity of 99.03%. These results were obtained using a data distribution strategy allocating 70% for training, 20% for testing, and 10% for validation. The Bi-TransNet architecture demonstrates unique versatility in effectively diagnosing an extensive range of neurological conditions within a unified framework, making it particularly suitable for diverse therapeutic applications. The model has achieved a significant developmental milestone through its robust generalization capability on previously unseen data, as evidenced by its efficacy on the DEAP dataset following training.},
        keywords = {Computer-aided Diagnosis; Electroencephalogram; Mental Illness; Recurrent Neural Network},
        month = {June},
        }