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

Deeksha M Shama; Punnisa Amornsirikul; Archana Venkataraman

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

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 Model	Spectral Domain				Spatial/Source Domain
Foundation Model	LRP $\dagger$	IG	DeepLIFT	DeepSHAP	LRP $\dagger$	IG	DeepLIFT	DeepSHAP
EEGNet	1.000 $\pm$ 0.000	0.942 $\pm$ 0.037	0.940 $\pm$ 0.039	1.000 $\pm$ 0.000	0.613 $\pm$ 0.043	0.655 $\pm$ 0.049	0.660 $\pm$ 0.052	0.600 $\pm$ 0.080
ABNet	1.000 $\pm$ 0.000	0.988 $\pm$ 0.008	0.978 $\pm$ 0.015	1.000 $\pm$ 0.000	0.655 $\pm$ 0.079	0.658 $\pm$ 0.088	0.655 $\pm$ 0.089	0.595 $\pm$ 0.088
Labram †	–	0.960 $\pm$ 0.022	0.972 $\pm$ 0.022	1.000 $\pm$ 0.000	–	0.692 $\pm$ 0.054	0.680 $\pm$ 0.063	0.638 $\pm$ 0.072
Cbramod †	–	0.993 $\pm$ 0.010	0.995 $\pm$ 0.006	1.000 $\pm$ 0.000	–	0.670 $\pm$ 0.057	0.638 $\pm$ 0.079	0.628 $\pm$ 0.090
LUNA †	–	0.945 $\pm$ 0.020	0.980 $\pm$ 0.013	0.998 $\pm$ 0.005	–	0.648 $\pm$ 0.093	0.645 $\pm$ 0.083	0.625 $\pm$ 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).

Dataset	Simulated	TUSZ	ACE
Input	8.137	4.163	3.239
Spectral	31.691	215.192	6.897
Spatial	13.395	7.714	8.981

TABLE II: mean-normalized range of EEG-PRISM attributions

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Model	Simulated	TUSZ	ACE
EEGNet	0.995 $\pm$ 0.006	0.878 $\pm$ 0.033	0.532 $\pm$ 0.023
ABNet	1.000 $\pm$ 0.000	0.878 $\pm$ 0.026	0.596 $\pm$ 0.036
Labram	1.000 $\pm$ 0.000	0.864 $\pm$ 0.039	0.660 $\pm$ 0.072
Cbramod	1.000 $\pm$ 0.000	0.868 $\pm$ 0.043	0.582 $\pm$ 0.062
LUNA	1.000 $\pm$ 0.000	0.878 $\pm$ 0.027	0.624 $\pm$ 0.043

Appendix

TABLE III: model accuracy in classification across datasets

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

EEG-Xplain: Decoding Neural Black-Boxes of EEG Foundation Models
Sep 14, 2026Hansong Ma, Junxiao WangGradient-Based Attribution Perturbation-Based Feature Attribution
Guangzhou University, China
What Do EEG Foundation Models Capture from Human Brain Signals?
May 12, 2026Ling Tang, Qian Chen, Jilin Mei +6Frequency-Domain Feature Learning Representation Probing
Shanghai Artificial Intelligence Laboratory · Shanghai Jiao Tong University · Tongji University +2
PRISM: Exploring Heterogeneous Pretrained EEG Foundation Model Transfer to Clinical Differential Diagnosis
Feb 28, 2026Jeet Bandhu Lahiri, Parshva Runwal, Arvasu Kulkarni +4EEG Foundation Models Transfer Learning
Indian Institute of Technology Mandi, India · NeuroDx, India

Sep 14, 2026J/K move · Enter open · S save

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

Hansong Ma, Junxiao Wang

Guangzhou University, China

Gradient-Based Attribution Perturbation-Based Feature Attribution Explainability Evaluation EEG Foundation Models

Open paper AskPDF arXiv

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.

Abstract

Neural Network Interpretability

Explore similar work

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

What Do EEG Foundation Models Capture from Human Brain Signals?

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