cs.LGAug 13, 2026

EEG-PRISM: Physiologically-Grounded Interpretability of Predictions by EEG Foundation Models

Authors: Deeksha M Shama, Punnisa Amornsirikul, Archana Venkataraman

Organizations: Department of Electrical and Computer Engineering at Johns Hopkins University, MD, USA 21218 · Johns Hopkins University

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Objective: Foundation models represent the next advancement in AI for EEG analysis; however current explainable AI techniques provide attribution scores in the time-channel input space, which is mismatched to clinical intuition about EEG. Thus, there is a critical need for a universal method that can extend the interpretability of any foundation model to alternative and physiologically relevant domains without modifying or retraining the underlying model. Methods: EEG-PRISM leverages linear transformations and established backpropagation rules to map time-channel attribution scores into alternative domains. We derive mappings to the frequency domain via an invertible DFT and to the source domain via an approximately invertible EEG generative model. We evaluate EEG-PRISM in simulated and real data, assessing recovery of ground-truth phenomena across domains with five foundation models and four AI explainers. Results: In simulation, EEG-PRISM achieves near-perfect spectral recovery and 69.2% spatial accuracy. In epilepsy, EEG-PRISM correctly determines that delta-theta activity is most salient and correctly localizes the seizure onset region with 50% accuracy. In autism, EEG-PRISM localizes the predictive delta-alpha biomarkers to frontal and temporal regions, consistent with prior work. Conclusion: EEG-PRISM is a theoretically-grounded post-hoc attribution method with accurate mapping into the spectral and spatial domains. It supports window-level analysis of transient events (e.g., seizures) and group-level identification of clinically relevant biomarkers (e.g., autism), thus advancing interpretable EEG foundation models. Significance: This work enables physiologically-grounded interpretation of EEG foundation models and supports clinically relevant insights such as event localization and biomarker identification.

Figures & tables

Fig. 1: EEG foundation models and their explainability using post-hoc tools operate in time-channel space, while clinically meaningful patterns lie in spectral and source domains that govern EEG generation.
Fig. 2: Overall pipeline of EEG-PRISM for explainability in clinically relevant domains beyond the input time-channel space. Top: End-to-end experimental setup, in which EEG-PRISM is applied to foundation models post-hoc to obtain attributions in the spectral and spatial domains. Bottom left: Transforming attributions to new domains within EEG-PRISM. Bottom right: Potential downstream applications.
Fig. 3: Simulated EEG data generation pipeline using SEREEGA. (1) Initialize the head model and select a source location, (2) Assign class-associated frequency signal, (3) Generate EEG from source activity using the leadfield matrix, and (4) Preprocess the EEG using standard methods.
Foundation ModelSpectral DomainSpatial/Source Domain
LRP †\daggerIGDeepLIFTDeepSHAPLRP †\daggerIGDeepLIFTDeepSHAP
EEGNet1.000 ±\pm 0.0000.942 ±\pm 0.0370.940 ±\pm 0.0391.000 ±\pm 0.0000.613 ±\pm 0.0430.655 ±\pm 0.0490.660 ±\pm 0.0520.600 ±\pm 0.080
ABNet1.000 ±\pm 0.0000.988 ±\pm 0.0080.978 ±\pm 0.0151.000 ±\pm 0.0000.655 ±\pm 0.0790.658 ±\pm 0.0880.655 ±\pm 0.0890.595 ±\pm 0.088
Labram †–0.960 ±\pm 0.0220.972 ±\pm 0.0221.000 ±\pm 0.000–0.692 ±\pm 0.0540.680 ±\pm 0.0630.638 ±\pm 0.072
Cbramod †–0.993 ±\pm 0.0100.995 ±\pm 0.0061.000 ±\pm 0.000–0.670 ±\pm 0.0570.638 ±\pm 0.0790.628 ±\pm 0.090
LUNA †–0.945 ±\pm 0.0200.980 ±\pm 0.0130.998 ±\pm 0.005–0.648 ±\pm 0.0930.645 ±\pm 0.0830.625 ±\pm 0.089
TABLE I: Accuracy of EEG-PRISM derived attributions vs ground-truth features in simulated dataset. Left: Comparison between the top spectral band and ground truth band. Right: Comparison between the center of mass location and ground truth source.
Fig. 4: EEG-PRISM results on TUSZ dataset. Top: Range of spectral attributions aggregated into five bands across all EEG windows in five models. Bottom: Confusion Matrices (CM) comparing region with maximum aggregated spatial attributions of EEG-PRISM vs ground truth seizure onset zone in clinical notes for five models. LA: Left Anterior; RA: Right Anterior; LP: Left Posterior; RP: Right Posterior; CS: Central/Subcortical
Fig. 5: EEG-PRISM results on ACE dataset grouped into ASD subjects (top), TDC subjects (middle), and respective differences across two groups (bottom). Spectral attributions are shown as box plots (left), and spatial attributions are mapped onto the 3D surface (right).
DatasetSimulatedTUSZACE
Input8.1374.1633.239
Spectral31.691215.1926.897
Spatial13.3957.7148.981
TABLE II: mean-normalized range of EEG-PRISM attributions
Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

ModelSimulatedTUSZACE
EEGNet0.995 ±\pm 0.0060.878 ±\pm 0.0330.532 ±\pm 0.023
ABNet1.000 ±\pm 0.0000.878 ±\pm 0.0260.596 ±\pm 0.036
Labram1.000 ±\pm 0.0000.864 ±\pm 0.0390.660 ±\pm 0.072
Cbramod1.000 ±\pm 0.0000.868 ±\pm 0.0430.582 ±\pm 0.062
LUNA1.000 ±\pm 0.0000.878 ±\pm 0.0270.624 ±\pm 0.043
Appendix
TABLE III: model accuracy in classification across datasets
More in this topic

Neural Network Interpretability

Neural network interpretability studies how models represent information and make predictions, and how to explain, test, or control their internal computations.

See all 330 papers

Explore similar work

CardsList
Sep 14, 2026

EEG-Xplain: Decoding Neural Black-Boxes of EEG Foundation Models

EEG foundation models such as BIOT, LaBraM, and EEGMamba have achieved remarkable performance in neural signal decoding, but their black-box nature limits clinical trust and neuroscientific validation. We propose a unified attribution framework for interpreting EEG foundation models across heterogeneous architectures. The framework integrates gradient-, perturbation-, and activation-based explanation methods to analyze model behavior in spatial, temporal, and frequency dimensions. Spatially, it identifies critical EEG channels and visualizes their distributions using topographic maps. Temporally, it highlights decision-relevant signal segments through attribution heatmaps. In the frequency domain, it quantifies the contributions of canonical EEG rhythms via spectral perturbation analysis. To assess explanation reliability, we introduce a population-level evaluation combining Area Over the Perturbation Curve (AOPC) and cross-method consistency analysis. The framework further leverages Large Language Models (LLMs) to transform structured attribution outputs into natural-language reports, bridging low-level neural representations and high-level semantic reasoning. Experiments on benchmark datasets, including Mumtaz2016 and TUAB, demonstrate that the generated explanations are consistent with established neurophysiological markers, validating meaningful neural representations while exposing potential dependencies on artifacts and spurious patterns. The proposed framework provides a standardized approach for evaluating the interpretability, reliability, and physiological plausibility of EEG foundation models.

Hansong Ma, Junxiao Wang

Guangzhou University, China

Gradient-Based AttributionPerturbation-Based Feature AttributionExplainability EvaluationEEG Foundation Models
AskPDF
May 12, 2026

What Do EEG Foundation Models Capture from Human Brain Signals?

Clinical electroencephalogram (EEG) analysis rests on a hand-crafted feature catalog refined over decades, e.g., band power, connectivity, complexity, and more. Modern EEG foundation models bypass this catalog, learn directly from raw signals via self-supervised pretraining, and match or outperform feature-engineered baselines on most clinical benchmarks. Whether the two representations align is an open question, which we decompose into three sub-questions: what does the model learn, what does the model use, and how much can be explained. We answer them with layer-wise ridge probing, LEACE-style cross-covariance subspace erasure, and a transparent classifier benchmarked against a random-feature baseline. The audit covers three foundation models (CSBrain, CBraMod, LaBraM), five clinical tasks (MDD, Stress, ISRUC-Sleep, TUSL, Siena), and a 6-family 63-feature lexicon. Of the 945945 (model, task, feature) units, 648648 (68.6%68.6\%) are representation-causal and 199199 (21.1%21.1\%) are encoded-only. Across tasks, 5050 features qualify as universal candidates with strong support (all three architectures RC) in two or more tasks. Frequency-domain features dominate, but the other five families each contribute substantial causal mass. Confirmed features recover, on average, 79.3%79.3\% of the foundation model's advantage over the random baseline, with a clean task gradient (MDD ≈0.99\approx 0.99 down to Stress ≈0.56\approx 0.56): tasks near ceiling are almost fully recovered by the lexicon, while harder tasks leave a non-trivial residual that pinpoints a concrete target for future concept discovery.

Ling Tang, Qian Chen, Jilin Mei +6

Shanghai Artificial Intelligence Laboratory · Shanghai Jiao Tong University · Tongji University +2

Frequency-Domain Feature LearningRepresentation ProbingEEG Foundation Models
AskPDF
Feb 28, 2026

PRISM: Exploring Heterogeneous Pretrained EEG Foundation Model Transfer to Clinical Differential Diagnosis

EEG foundation models are typically pretrained on narrow-source clinical archives and evaluated on benchmarks from the same ecosystem, leaving unclear whether representations encode neural physiology or recording-distribution artifacts. We introduce PRISM (Population Representative Invariant Signal Model), a masked autoencoder ablated along two axes -- pretraining population and downstream adaptation -- with architecture and preprocessing fixed. We compare a narrow-source EU/US corpus (TUH + PhysioNet) against a geographically diverse pool augmented with multi-center South Asian clinical recordings across multiple EEG systems. Three findings emerge. First, narrow-source pretraining yields stronger linear probes on distribution-matched benchmarks, while diverse pretraining produces more adaptable representations under fine-tuning -- a trade-off invisible under single-protocol evaluation. Trained on three source corpora, PRISM matches or outperforms REVE (92 datasets, 60,000+ hours) on the majority of tasks, demonstrating that targeted diversity can substitute for indiscriminate scale and that dataset count is a confounding variable in model comparison. Second, on a clinically challenging and previously untested task -- distinguishing epilepsy from diagnostic mimickers via interictal EEG -- the diverse checkpoint outperforms the narrow-source checkpoint by +12.3 pp balanced accuracy, the largest gap across all evaluations. Third, systematic inconsistencies between EEG-Bench and EEG-FM-Bench reverse model rankings on identical datasets by up to 24 pp; we identify six concrete sources including split construction, checkpoint selection, segment length, and normalization, showing these factors compound non-additively.

Jeet Bandhu Lahiri, Parshva Runwal, Arvasu Kulkarni +4

Indian Institute of Technology Mandi, India · NeuroDx, India

EEG Foundation ModelsTransfer Learning
AskPDF