NEUROTWIN: Multimodal Neuroimaging Digital Twin Laboratory
Computational digital twin framework reconstructing cortical anatomy, tractography, and connectomics
Problem Context & System Scope
NEUROTWIN is a scientific computational laboratory platform that ingests raw and preprocessed multimodal neuroimaging datasets (MRI, fMRI, DTI) to construct an interactive, stateful 3D computational model.
Unlike purely visual mesh viewers, NEUROTWIN maintains mathematical representations of cortical geometry, microstructural diffusion parameters, 3D white-matter streamlines, and structural connectivity matrices.
Includes a lesion simulation engine that calculates topological degradation (global efficiency loss, severed tract bundles, and downstream disconnected functional regions) in real time.
Technical Breakdown & Pipeline Engineering
Cortical Parcellation & 3D Geometry Pipeline
FreeSurfer surface ingestion and anatomical region mapping.
- Imports FreeSurfer fsaverage5 standard cortical pial surface with 20,482 vertices.
- Accurate mapping across Harvard-Oxford 59 Regions of Interest (ROIs) with physical MNI centroid coordinates.
- Real-time vertex-level shader projection for regional activity maps.
Diffusion Tensor Imaging & Tractography Engine
Stejskal-Tanner tensor reconstruction and streamline propagation.
- Parses 4D DWI NIfTI, .bval, and .bvec files to compute diffusion tensors (Dxx to Dzz).
- Derives quantitative microstructural biomarkers: Fractional Anisotropy (FA), Mean Diffusivity (MD), Radial Diffusivity (RD).
- Generates 3D white-matter tractography bundles (Corpus Callosum, Longitudinal Fasciculi) with standard directional RGB color mapping (R=|dx|, G=|dy|, B=|dz|).
Connectomics & Structural Lesion Simulator
Graph-theoretic topological analysis and disruption forecasting.
- Constructs 59x59 structural connectivity matrices normalized by geometric volume.
- Supports multimodal fusion schemes (Linear Hybrid, Anatomically Constrained).
- Provides interactive lesion simulation: computes topological damage, global routing efficiency drop %, and impacted neural pathways.
Engineering Hardships & Resolution
Implemented compressed line-strip geometry buffers with directional color attributes packed into compact WebGL vertex arrays.
Standardized on DIPY-compliant tensor estimation pipelines with explicit b-value thresholds and robust gradient table validations.
Quantitative Results & Impact
- Engineered comprehensive simulation framework supporting multi-modal MRI/fMRI/DTI analysis.
- Demonstrated real-time interactive lesion topological impact calculations in computational neuroscience.