FlexiBrain: New AI Framework Processes Brain Imaging Data Without Destructive Preprocessing
Researchers have developed FlexiBrain, a machine learning framework that processes fMRI brain scans directly in their native resolution without requiring standardized preprocessing. The method uses a Mamba-JEPA architecture with dynamic patch resizing to handle data from diverse sources with different spatial and temporal resolutions. This approach preserves anatomical detail, reduces computational costs, and outperforms existing methods by up to 12 percentage points on downstream neuroscience tasks.
FlexiBrain addresses a fundamental challenge in neuroscience AI: the heterogeneity of fMRI data collected from different sources with varying resolutions. Traditional approaches require lengthy preprocessing pipelines that standardize data to uniform formats, a process that can degrade subject-specific anatomical information and consume significant computational resources. The new framework bypasses this standardization by defining patch sizes in real-world physical units and dynamically resizing patches to accommodate native resolution variations. Built on an efficient Mamba-JEPA backbone designed to model high-dimensional 4D fMRI signals, FlexiBrain was tested across five diverse downstream neuroscience tasks and consistently outperformed recent state-of-the-art methods. The framework functions as a plug-in module that substantially reduces preprocessing overhead while accelerating development of robust voxel-level fMRI foundation models, with code made publicly available.
What different sources said
- arXiv cs.LGCenter
FlexiBrain: Resolution-Agnostic Voxel-Level Encoding for Native fMRI
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.