Abstract
Schizophrenia is a psychiatric disorder that presents significant diagnostic challenges due to its complex neurophysiological characteristics. This paper investigates the potential of Large Language Models (LLMs), such as OpenAI's GPT-4 and GPT-o1, in detecting schizophrenia through electroencephalography (EEG) analysis. Using the LMSU public ScZ EEG dataset, we conducted a series of experiments involving different types of input data, including raw EEG signals, frequency band summaries, and graphical representations of brain activity. Our findings demonstrate that LLMs can accurately classify schizophrenic and healthy individuals while offering interpretable, clinically relevant insights aligned with established EEG markers. By integrating these models into the diagnostic workflow, we explore the concept of Symbiotic AI, where LLMs act as cognitive collaborators, enhancing clinicians' ability to analyze complex data efficiently and transparently. This approach not only improves diagnostic accuracy but also facilitates real-time decision-making, paving the way for earlier and more precise detection of schizophrenia in clinical settings.