Inpainting the Neural Picture: Inferring Unrecorded Brain Area Dynamics from Multi-Animal Datasets
🤖 AI Summary
Current single-animal electrophysiological experiments cannot simultaneously record across the entire brain, hindering whole-brain dynamic modeling and cross-regional interaction analysis. To address this limitation, we propose NeuroPaint—the first method to adapt the masked autoencoder framework for cross-animal neural data imputation. Leveraging partially overlapping brain coverage across multiple Neuropixels recordings from different animals, NeuroPaint learns a shared latent dynamical structure to reliably infer unrecorded regional activity. Trained on a hybrid dataset combining large-scale real and synthetic neural recordings, NeuroPaint achieves high-fidelity reconstruction of cross-regional dynamics in both domains. Its core contribution is breaking the constraint of single-animal experiments by enabling unsupervised, cross-modal neural activity completion via multi-animal data fusion—establishing a novel paradigm for whole-brain dynamical modeling and multi-region causal inference.
Application Category
📝 Abstract
Characterizing interactions between brain areas is a fundamental goal of systems neuroscience. While such analyses are possible when areas are recorded simultaneously, it is rare to observe all combinations of areas of interest within a single animal or recording session. How can we leverage multi-animal datasets to better understand multi-area interactions? Building on recent progress in large-scale, multi-animal models, we introduce NeuroPaint, a masked autoencoding approach for inferring the dynamics of unrecorded brain areas. By training across animals with overlapping subsets of recorded areas, NeuroPaint learns to reconstruct activity in missing areas based on shared structure across individuals. We train and evaluate our approach on synthetic data and two multi-animal, multi-area Neuropixels datasets. Our results demonstrate that models trained across animals with partial observations can successfully in-paint the dynamics of unrecorded areas, enabling multi-area analyses that transcend the limitations of any single experiment.
Problem
Research questions and friction points this paper is trying to address.
Inferring unrecorded brain area dynamics from multi-animal datasets
Leveraging multi-animal data to understand multi-area brain interactions
Reconstructing missing brain activity using shared structure across individuals
Innovation
Methods, ideas, or system contributions that make the work stand out.
Masked autoencoding reconstructs unrecorded brain area dynamics
Training across animals with overlapping recorded areas
Inferring missing neural activity using shared cross-individual structure
🔎 Similar Papers
No similar papers found.
