New BridgeMIL AI framework boosts EEG diagnosis accuracy

A team of researchers from Tsinghua University and collaborators have introduced BridgeMIL, a novel two-stage framework that significantly improves EEG-based disease diagnosis by addressing a fundamental flaw in traditional approaches.

Current pipelines segment EEG recordings into short instances but inherit subject-level labels for each segment, which assumes all instances provide equally reliable diagnostic evidence. BridgeMIL decouples instance representation learning from subject-level supervision through a two-stage process: Stage 1 pretrains the encoder without inherited labels by aligning temporally nearby windows and within-subject sub-bags, using variance and covariance regularization to prevent collapse and reduce redundancy. Stage 2 transfers the encoder to an attention-based MIL aggregator, applying supervision only to subject predictions while limiting representation drift. Across three EEG disease datasets and five backbones, BridgeMIL achieved the highest mean accuracy in 14 of 15 settings, reaching 76.57% overall—4.28 percentage points higher than the strongest baseline. Analyses revealed substantial variation in label reliability across instances and greater performance sensitivity to subject scarcity than instance scarcity.