Unsupervised learning of mapping between brain lesions and behavior

Causal Feature Learning (CFL) uncovers lesion-behavior relationships without predefined regions, outperforming traditional methods in simulated and real datasets.

• Why it matters: Understanding how specific brain lesions relate to behavioral outcomes is crucial for neuroscience but is hindered by the challenge of choosing the right analysis granularity and predefined regions.
• What they did: The authors developed CFL, a data-driven approach that simultaneously learns the appropriate level of analysis and the lesion-behavior relation, tested on simulated data and large human lesion datasets.
• The result: CFL successfully recovers known and novel lesion-behavior maps, enabling more accurate and flexible identification of brain-behavior relationships and revealing new groupings of behavioral outcomes.