BrainWave: A Brain Signal Foundation Model for Clinical Applications

TL;DR

BrainWave introduces the first brain-signal foundation model trained on both invasive and non-invasive recordings, enabling robust, generalizable representations through a scale-alignment layer and masked time-frequency pretraining. The model demonstrates state-of-the-art performance across cross-domain disease diagnosis, few-shot learning, and real-world clinical tasks such as SOZ localization and Alzheimer's biomarker prediction, with strong zero-shot transfer across hospitals and seizure subtypes. A key finding is that joint pretraining on EEG and iEEG yields richer representations and better generalization than single-modality baselines, implying broad potential for multimodal biosignal AI in medicine. Open-sourcing BrainWave could accelerate AI-assisted brain disorder research and clinical decision support, while future work may extend the approach to additional modalities like MRI and EMG.

Abstract

Neural electrical activity is fundamental to brain function, underlying a range of cognitive and behavioral processes, including movement, perception, decision-making, and consciousness. Abnormal patterns of neural signaling often indicate the presence of underlying brain diseases. The variability among individuals, the diverse array of clinical symptoms from various brain disorders, and the limited availability of diagnostic classifications, have posed significant barriers to formulating reliable model of neural signals for diverse application contexts. Here, we present BrainWave, the first foundation model for both invasive and non-invasive neural recordings, pretrained on more than 40,000 hours of electrical brain recordings (13.79 TB of data) from approximately 16,000 individuals. Our analysis show that BrainWave outperforms all other competing models and consistently achieves state-of-the-art performance in the diagnosis and identification of neurological disorders. We also demonstrate robust capabilities of BrainWave in enabling zero-shot transfer learning across varying recording conditions and brain diseases, as well as few-shot classification without fine-tuning, suggesting that BrainWave learns highly generalizable representations of neural signals. We hence believe that open-sourcing BrainWave will facilitate a wide range of clinical applications in medicine, paving the way for AI-driven approaches to investigate brain disorders and advance neuroscience research.

Figures (12)

• Figure 1: Overview of BrainWave. a, Data curation for pretraining BrainWave. The pretraining corpus contains both invasive and non-invasive brain recordings collected from diverse healthcare scenarios. b, The pretraining pipeline of BrainWave. BrainWave is pretrained on more than 3 billion signal patches using a masked modeling strategy. c, The evaluation tasks consist of few-shot classification and cross-domain evaluation. We conduct few-shot classification with a prototypical network in which the we directly compare the representations of the queries with class prototypes. We perform three different levels of cross-domain analysis: cross-subject, cross-hospital, and cross-subtype. d, The overall results of BrainWave compared to other pretrained models. BrainWave outperforms other models across all the 24 experiments, with significant improvement ($p < 0.001$) in 20 of them.
• Figure 2: Performance of cross-domain evaluation.a-j, Bar plots comparing the AUROC scores of BrainWave and competing models on cross-subject tasks. Each experiment is conducted with $n$-fold cross validation ($n$ is the number of subject groups), where we repeat five runs for each fold. k,l, Bar plots comparing the AUROC scores of BrainWave and competing models on cross-hospital tasks. m,n, Bar plots comparing the AUROC scores of BrainWave and competing models on cross-subtype tasks. k-n, The source dataset is served as the training set and the target dataset is served as the evaluation set. In each experiment, we repeat five runs. a-n, Data are mean $\pm$ SD. The listed $p$ value indicates the significance for BrainWave outperforming the best comparison model, with the two-sided $t$-test.
• Figure 3: Performance and analysis of few-shot classification.a-j, Box plots comparing the AUROC scores of BrainWave and competing models on few-shot classification. We conduct $n$-fold cross validation for each experiment and repeat five runs per fold. We perform 3-shot and 8-shot classification for each task. k, t-SNE (t-distributed Stochastic Neighbor Embedding) plots of the pretrained representations on Absence-16 generated from BrainWave and other pretrained encoders. Each model contains four subplots, with each subplot generated by randomly sampling a portion of the original dataset.
• Figure 4: Clinical tasks on epilepsy and AD. a, Channel-level epileptic waveform detection on 4 patients with DRE. Sensitivity/Specificity and AUROC/AP are reported. b, The process of pinpointing channels with frequent epileptic discharges and repeated involvement as seizure onset sites. We can quickly quantify these metrics for each channel from model predictions. c-g, Predictions of amyloid-beta deposition and a series of clinical scale scores from AD patients. Each experiment is conducted with 5-fold cross validation, where we repeat five runs for each fold. c, Amyloid-beta deposition prediction. d, MMSE score prediction. We divide it into 4 discrete ranges: 24-30, 21-23, 10-20 and 0-9. e, MoCA-B score prediction. We divide it into 4 discrete ranges: 26-30, 18-25, 10-17 and 0-9. f, ROCF score prediction. We divide it into 5 discrete ranges: 33-36, 24-32, 18-23, 12-17 and 0-11. g, PSQI score prediction. We divide it into 4 discrete ranges: 0-5, 6-10, 11-15 and 16-21.
• Figure 5: Analysis of joint pretraining. a, Scatter plots comparing the AUROC and BACC scores of BrainWave, BrainWave-EEG and BrainWave-iEEG on cross-subject tasks. b, Bar plots comparing the AUROC and BACC scores of BrainWave and BrainWave-iEEG on cross-hospital tasks. c, Bar plots comparing the AUROC and BACC scores of BrainWave and BrainWave-EEG on cross-subtype tasks. b,c, The source dataset is served as the training set and the target dataset is served as the evaluation set. In each experiment, we repeat five runs. Data are mean $\pm$ SD. The listed $p$ value indicates the significance for BrainWave outperforming the best comparison model, with the two-sided $t$-test. d, Box plots comparing the AUROC scores of BrainWave, BrainWave-EEG and BrainWave-iEEG on few-shot classification. We perform 3-shot and 8-shot classification for each dataset. e, Average improvement of BrainWave over BrainWave-EEG and BrainWave-iEEG on cross-domain evaluation and few-shot classification. We first calculate the relative improvement for each experiment and then compute the average of them. f, Bar plots comparing the AUROC and BACC scores of BrainWave, BrainWave-EEG and BrainWave-iEEG on out-of-domain recording type evaluation. We collect a ECG dataset (Apnea-ECG) with a sleep apnea detection task. Data are mean $\pm$ SD. a,d,f, Each experiment is conducted with $n$-fold cross validation ($n$ is the number of subject groups), where we repeat five runs for each fold.