<p>Schizophrenia is a complex neuropsychiatric disorder that is not easy to diagnose because of the reason that the variability of symptoms and insufficient definitive biomarkers. Recent advancements in machine learning (ML), particularly in the study related to Electroencephalogram (EEG) signals, offer promising avenues for early diagnosis. This study proposes an automated approach for schizophrenia detection using EEG spectrograms, leveraging Transfer Learning (TL) and Self-Distillation (SD) methods. Transfer learning is employed to adapt pre-trained models from other domains to the task of schizophrenia detection, thereby improving classification accuracy despite limited labelled data. Additionally, self-distillation is incorporated to further enhance model performance by utilizing a teacher-student framework. In this framework, the student network is made to learn from both the ground-truth labels and the teacher model’s predictions. The method proposed here uses a publicly available EEG dataset. It gives a remarkable improvement in classification accuracy and robustness in comparison of the traditional used ML techniques. This research article shows the potential of combining transfer learning and self-distillation to advance automated schizophrenia detection and provide a scalable solution for clinical practice.</p>