BrainMIT

The aim of the BrainMIT project is to develop a virtual biopsy tool capable of producing in vivo a digital twin of the brain tissue microstructure of any healthy or pathological individual, based on diffusion-weighted MRI data. Such an in vivo exploration tool, enabling navigation in brain tissue from the millimetric scale of MRI to the microscopic scale of cells, requires a multi-disciplinary team covering the diverse expertise needed to successfully complete the project. The BrainMIT consortium therefore involves:

1) clinicians (neurosurgeons and neuroanatomists) to gain access to post-mortem human brains and provide the clinical expertise needed to build a tool to improve their daily practice,

2) recognized experts in the field of neuroimaging, particularly magnetic resonance imaging (MRI) and microscopic imaging of stained histological sections, and

3) recognized experts in numerical simulation, artificial intelligence (AI) and high-performance computing (HPC).

The project will be organized around four tasks:

1) After an initial stage of obtaining reference data acquired post-mortem by light microscopy on histological sections from human brain anatomical parts stained by immunohistochemistry, the key descriptors characterizing the microstructure of brain tissue will be extracted using artificial intelligence (AI) methods for segmenting individual cells.

2) The microstructural a priori derived from this ground truth will then be used to create large-scale numerical simulations of the cellular organization of brain tissue, as well as their signatures in diffusion MRI using Monte-Carlo approaches implemented on national supercomputers.

3) In a third stage, the Big Data thus produced will enable AI experts to design a computational model for producing digital twins of brain tissue microstructure in humans from a millimetric diffusion MRI dataset acquired in vivo at an individual scale.

4) Finally, the BrainMIT tool will be validated using a cohort of elderly subjects enrolled in a body donation program, for whom the microstructural data inferred from in vivo MRI will be validated by comparison with histology after their death. The validated tool will then exploit MRI and microscopy data acquired in the PEPR-SN BrainDeepPhenotyping collaborative project to build digital twins of the brains of patients with neurodegenerative and neurovascular pathologies.

The BrainMIT project will revolutionize the characterization of cellular damage in multiple brain pathologies. It will improve clinical practice and patient comfort by offering a completely non invasive virtual biopsy, facilitating diagnosis and longitudinal therapeutic follow-up. Finally, the possibility of accessing such digital twins of an individual’s brain from simple MRI acquisitions will enable in vivo characterization of the cellular organization of functional brain networks in the same individual, representing a major advance in the field of cognitive neuroscience.