Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study benchmarks six EEG foundation models across five clinical tasks and four datasets to assess their robustness and transferability in clinical decoding. The findings reveal that while the REVE model shows promise in cross-subject ictal detection, its performance varies significantly depending on the evaluation unit and dataset characteristics, with classical features often outperforming it in other contexts. Notably, dataset identity can be easily decoded from frozen embeddings, highlighting the potential pitfalls of relying solely on pretrained models for clinical applications.
EEG foundation models can mislead clinical interpretations, as dataset identity can be decoded with near-perfect accuracy while clinical tasks often yield subpar results.
Pretrained EEG foundation models are increasingly proposed for clinical decoding, but their transfer across populations and robustness to negative controls remain unclear. We benchmark six models (LaBraM, EEGMamba, CBraMod, REVE, BENDR, and BIOT) on five clinical tasks across four datasets using frozen linear probes with leave-one-subject-out, subject-grouped, or explicitly identified recording-level splits. Selected REVE findings are tested against random initialisation, random features, label permutation, scrambled-label fine-tuning, and projection sensitivity. On Korean dementia (CAUEEG, three-way), frozen REVE reaches 0.568 AUROC versus 0.769 for classical features; the ordering persists on a patient-disjoint held-out split (0.565 versus 0.768). Dataset identity is readily decoded from frozen embeddings (AUROC 1.000 at PCA-50; 0.9998 after band restriction and per-epoch z-scoring), whereas the same PCA-50 pipeline decodes Korean diagnosis at 0.528. A randomly initialised encoder also outperforms pretrained REVE on this task (0.659 versus 0.570). On Alzheimer's disease, Gaussian random projection and PCA of the same pretrained embeddings perform similarly, and classical features nominally exceed REVE at the subject level. The clearest controlled positive is cross-subject ictal detection on CHB-MIT (n=23), where REVE achieves 0.793 AUROC, 9.2 percentage points above a randomly initialised encoder. These results show that EEG foundation-model conclusions depend strongly on evaluation unit, dataset shift, comparator strength, and targeted controls.