CortexLab: Open-source toolkit for brain-alignment benchmarking built on TRIBE v2

[SID_Rajhans]

Hi everyone,

I’ve built CortexLab, an open-source Python toolkit that extends Meta’s TRIBE v2 for computational neuroscience research.

What it does:

TRIBE v2 predicts fMRI brain activation from video, audio, and text. CortexLab adds the analysis layer:

• Brain-Alignment Benchmark: Score any AI model’s representations against brain responses using RSA, CKA, and Procrustes with permutation tests, bootstrap CIs, and FDR correction

• Cognitive Load Scoring: Map predicted activations onto 4 neurocognitive dimensions (visual, auditory, language, executive) using HCP MMP1.0 ROIs

• Temporal Dynamics: Peak response latency, lag correlations, sustained/transient decomposition

• ROI Connectivity: Functional connectivity matrices, network clustering, modularity, centrality

• 3D Brain Viewer: Interactive fsaverage brain with activation overlays

• Live Inference: Real-time brain prediction from webcam or video

Try it:

• Live demo (no install): CortexLab Dashboard - a Hugging Face Space by SID2000

• GitHub: GitHub - siddhant-rajhans/cortexlab: Enhanced multimodal fMRI brain encoding toolkit built on Meta's TRIBE v2. Features streaming inference, brain-alignment benchmarking, cognitive load scoring, and more. · GitHub

• pip install cortexlab-toolkit

89 tests, CC BY-NC 4.0, contributions welcome.

I’d especially appreciate feedback on:

• The ROI-to-cognitive-dimension mapping methodology

• Additional alignment metrics worth implementing

• Validation approaches for the cognitive load scorer

Thanks!

Tags

fmri, python, brain-encoding, neuroscience, open-source