Do Brain MRI Foundation Models Encode Epilepsy-Relevant Concepts? A Confound-Aware Linear Probing Study on Focal Cortical Dysplasia

Linear probing is the standard cheap test of whether a frozen foundation model has ``already learned'' a clinical concept, and its result is almost always read against chance. We argue that chance alone is an insufficient reference point for neuroimaging, and show why on a case where the stakes are concrete. We probe frozen BrainIAC embeddings for focal cortical dysplasia (FCD) type II, a surgically treatable cause of drug-resistant epilepsy, using all 170 subjects of the open Bonn FCD-II cohort. The lesion-presence probe reaches AUC $0.694 \pm 0.028$, which exceeds chance but not the confound-only baseline. The same embeddings decode acquisition era at AUC $0.986$, era is coupled to the clinical label, and a classifier using only the one-bit era tag---with no image at all---already reaches AUC $0.723$. Restricting the probe to a single acquisition era, thereby eliminating between-era variation without modelling it, collapses performance to $0.566$ with a 95\% CI of $[0.472, 0.631]$ and a subject-level permutation $p = 0.13$. Hemispheric lateralization never leaves chance. We argue that clinical concept-probing studies should routinely report confound probes and explicit confound-only baselines, and that a dataset can be the best available for a disease and still be unable to support the claim that a model has encoded it.