The field of brain-computer interfaces (BCIs) and neuroimaging is rapidly advancing, with a focus on developing more accurate and efficient models for brain disease localization, diagnosis, and prediction. Recent developments have highlighted the potential of self-calibrating BCIs, which can recover mental targets without labeled data, and foundation models for fMRI analysis, which can improve reproducibility and transferability across diverse applications. Multimodal graph learning frameworks have also shown promise in precise and computationally efficient identification of brain regions affected by neurodegenerative diseases. Furthermore, hybrid deep learning models and multimodal foundation models have demonstrated improved performance in multiclass brain disease classification and early prediction of multiple sclerosis disability progression. Noteworthy papers include:

• Self-Calibrating BCIs: Ranking and Recovery of Mental Targets Without Labels, which presents a framework for recovering unknown mental targets without labeled data.
• Towards a general-purpose foundation model for fMRI analysis, which introduces a generalizable framework that enables efficient knowledge transfer across diverse applications.
• BrainMAP: Multimodal Graph Learning For Efficient Brain Disease Localization, which presents a novel multimodal graph learning framework for precise and computationally efficient identification of brain regions affected by neurodegenerative diseases.