The field of neuropsychiatric disorder diagnosis is moving towards the development of more interpretable and reliable diagnostic models. Recent studies have focused on leveraging graph neural networks and information bottleneck principles to identify informative brain regions and functional connectivity patterns. These approaches have shown promise in improving diagnostic accuracy and providing clinically meaningful explanations. The use of concept-guided graph neural networks and large language models has also been explored, enabling the generation of interpretable functional connectivity concepts and disorder-specific connectivity patterns. Furthermore, research has highlighted the importance of anonymization and privacy protection in the analysis of EEG data, as it can reveal sensitive personal health information. Noteworthy papers include:

• BrainIB++: Leveraging Graph Neural Networks and Information Bottleneck for Functional Brain Biomarkers in Schizophrenia, which introduced an end-to-end graph neural network framework for identifying informative brain regions.
• Interpretable Neuropsychiatric Diagnosis via Concept-Guided Graph Neural Networks, which proposed a concept-based diagnosis framework that leverages large language models and neurobiological domain knowledge to generate interpretable functional connectivity concepts.