A Multimodal Brain-Signal Foundation Model (MBFM) for unified brain analysis and disease diagnosis

Project summary

This project aims to develop a multimodal brain-signal foundation model (MBFM) that can integrate diverse brain data types such as spiking activity, EEG, and imaging. Current models struggle with multimodal integration due to fragmented and small datasets. By extending existing foundation-model architectures, we will unify diverse datasets for better brain-signal analysis, improving applications like brain-computer interfaces (BCIs) and neurodegenerative-disease diagnosis. Though ultimately targeting humans, as a proof of concept, the MBFM will be trained on a wealth of publicly available rodent datasets and internal clinical data, supported by our industrial expert’s –Groq– AI infrastructure and know-how for efficient multimodal processing.