Despite the remarkable advancements in neuroscience and related fields, we still cannot fully explain the complexities of the human brain. Foundation models hold promise, but so far, these applications have focused only on single modalities of data, such as MRI, or one of its functions, such as vision. Our team plans to build on these early efforts by integrating four interconnected domains essential to our health: psychological, biological, environmental, and physical. Our first step is to unify large-scale datasets across various facets of human health. We will train domain-specific models end-to-end to develop a comprehensive Brain World Model. With these insights, we hope to enable more precise diagnostics and personalized patient care in clinical settings.
Multimodal Brain Imaging Pipelines (sMRI, fMRI, DTI)
This project provides an integrated suite of processing pipelines for structural MRI (sMRI), functional MRI (fMRI), and diffusion MRI (DTI). Each pipeline is modular, reproducible, and equipped with built-in quality control (QC) tools, enabling robust and comparable analyses across large-scale neuroimaging studies.
Modalities & Capabilities
| Modality | Core Functionality | Highlights |
|---|---|---|
| sMRI | Brain extraction (skull stripping), registration to MNI space (rigid + affine), post-processing (N4 bias correction, masked Z-score normalization, cropping), QC report with visualizations and Dice-based, and CSV metadata output. | Pipeline |
| fMRI | Built on fMRIPrep, with flexible configuration for GPU/CPU parallelization. Supports connectivity matrix generation and provides QC reports, including the number of volumes, Framewise Displacement (FD), and High-Motion Ratio (percentage of high-motion frames). Includes visual QC for connectivity matrices, plus separate fMRIPrep reports embedded in QC. | Pipeline |
| DTI | Susceptibility distortion correction (TOPUP), eddy current & motion correction (EDDY), within-session alignment/merge, registration to MNI, diffusion tensor fitting via DIPY (FA, MD, RD, AD, eigenvectors, tensor components), QC visualizations and summary tables (CSV/JSON/HTML), plus Dice-based QC for MNI registration. | Pipeline |