Interpretable Fuzzy Modeling Reveals Population-Level Representation Differences in P300 Brain Computer Interfaces Across Neurodivergent and Neurotypical Cohorts

Xiaowei Jiang; Sudong Shang; Adrian Wilkinson; Michael L. Platt; Da Xiao; Bening Cao; Thomas Do

eess.SPApr 4, 2026

Interpretable Fuzzy Modeling Reveals Population-Level Representation Differences in P300 Brain Computer Interfaces Across Neurodivergent and Neurotypical Cohorts

Authors: Xiaowei Jiang, Sudong Shang, Adrian Wilkinson, Michael L. Platt, Da Xiao, Bening Cao, Thomas Do

Organizations: Computational Intelligence and Brain-Computer Interfaces lab, Australian AI Institute, Faculty of Engineering and Information Technology, University of Technology Sydney, 2007, Australia · Department of Management, Griffith Business School, 4215, Australia · Department of Neuroscience, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, 19104, USA · Marketing Department, Wharton School of Business, University of Pennsylvania, Philadelphia, Pennsylvania, 19104, USA ·

Abstract

P300-based brain-computer interfaces (BCIs) are widely used for communication, but population heterogeneity may alter the neural patterns available for decoding. Prior work has mainly examined such differences at the signal or performance level, while the representation structure learned by the decoder remains underexplored. In this study, we propose an interpretable fuzzy spatiotemporal framework for P300 classification and use it to analyze population-level differences across amyotrophic lateral sclerosis (ALS), autism (AUT), and neurotypical (NT) cohorts. The model employs spatial and temporal fuzzy filters with learnable prototypes, enabling both classification and reconstruction of cohort-specific fuzzy centers. Experiments were conducted on ALS and NT subsets from bigP3BCI and on the BCIAUT-P300 benchmark in a within-subject setting. The proposed model achieved competitive performance against multiple deep learning baselines. More importantly, the reconstructed fuzzy centers revealed systematic cohort-dependent differences in waveform morphology and representation geometry. Point-wise statistical analysis identified significant temporal differences between cohorts, including intervals overlapping with the canonical P300 window, and low-dimensional embeddings showed partially separated cohort-specific prototype organizations. These results suggest that population heterogeneity in P300-BCI is reflected not only in decoding performance but also in the discriminative structure learned by the model. The proposed framework provides an interpretable route toward population-aware P300-BCI analysis and design.

Figures & tables

Fig. 1: Overall architecture of the proposed interpretable fuzzy P300 decoder. The model consists of a spatial fuzzy filter, a temporal fuzzy filter, and a classification MLP head. Given an input EEG epoch of shape $(B,D,T)$ , where $B$ denotes batch size, $D$ denotes the number of channels, and $T$ denotes the number of time points, the spatial fuzzy filter first performs rule-guided aggregation across the channel dimension using $R_{s}$ learnable fuzzy centers. The resulting representation is then processed by the temporal fuzzy filter, which applies an analogous fuzzy inference mechanism across the temporal dimension using $R_{t}$ learnable fuzzy centers.

Cohort	Study	Model	Accuracy	F1	Precision	Recall
ALS	Study L	Fuzzy	86.81 (2.69)	82.53 (3.71)	80.59 (4.92)	86.81 (2.69)
		Conformer-BIGP3BCI	86.02 (3.77)	81.51 (4.50)	77.88 (5.66)*	86.02 (3.77)
		DeepConvNet-BIGP3BCI	86.69 (4.21)	86.40 (4.55)**	86.49 (5.07)**	86.69 (4.21)
		EEGNet-BIGP3BCI	87.87 (3.19)*	85.67 (4.36)*	85.29 (4.70)*	87.87 (3.19)*
		ShallowConvNet-BIGP3BCI	88.60 (3.40)*	86.97 (4.02)**	86.40 (4.39)**	88.60 (3.40)*
ALS	Study N	Fuzzy	88.46 (1.44)	84.59 (2.25)	83.00 (3.79)	88.46 (1.44)

TABLE I: Performance comparison between the proposed fuzzy model and baseline deep learning models across cohorts and studies. Results are reported as mean (standard deviation). Significance markers for baseline models indicate paired comparisons against the fuzzy model: *

p<0.05

, **

p<0.01

, ***

p<0.001

Fig. 2: Interpretable analysis of reconstructed fuzzy centers across cohorts. Left: reconstructed center waveforms for five representative rules learned from the ALS, AUT, and NT cohorts. Horizontal colored bars above each panel indicate time intervals showing significant effects from point-wise one-way ANOVA and post-hoc pairwise comparisons between ALS and NT, and between AUT and NT. The black vertical line marks stimulus onset, and the red window highlights the approximate P300-related interval. Right: low-dimensional embeddings of the reconstructed centers for each rule, showing cohort-specific organization in representation space. Across rules, the learned centers exhibit systematic waveform and geometric differences between ALS, AUT, and NT, supporting the presence of population-level representational heterogeneity.

Brain-Computer Interfaces

Brain-computer interfaces translate neural activity into commands or feedback for communication, movement, rehabilitation, and sensory restoration.

See all 130 papers

Sparse Bayesian Modeling of EEG Channel Interactions Improves P300 Brain-Computer Interface Performance
Feb 19, 2026Guoxuan Ma, Yuan Zhong, Moyan Li +2Brain-Computer Interfaces EEG Decoding
Department of Biostatistics, University of Michigan, Ann Arbor MI 48109
When More Is Not Better: Component Anti-Synergy in a P300 Speller
Sep 10, 2026Lucas Yang, Rui Liu, Fusheng WangBrain-Computer Interfaces EEG Decoding
Parkland High School Allentown, PA, USA · Dept. of Computer Science Stony Brook University Stony Brook, NY, USA · Dept. of Computer Science Dept. of Biomedical Informatics Stony Brook University Stony Brook, NY, USA
Explainability of Recurrent Neural Networks for Enhancing P300-based Brain-Computer Interfaces
May 11, 2026Christian Oliva, Vinicio Changoluisa, Francisco B Rodríguez +1Brain-Computer Interfaces EEG Decoding
Grupo de Neurocomputación Biológica, Departamento de Ingeniería Informática, Escuela Politécnica Superior, Universidad Autónoma de Madrid, c/ Francisco Tomas y Valiente, 11, Madrid, 28049, Madrid, Spain · Grupo de Investigación en Electrónica y Telemática, Universidad Politécnica Salesiana, Campus el Girón, Ave 12 de Octubre 24-22, Quito, 170143, Quito, Ecuador

Feb 19, 2026J/K move · Enter open · S save

Sparse Bayesian Modeling of EEG Channel Interactions Improves P300 Brain-Computer Interface Performance

Guoxuan Ma, Yuan Zhong, Moyan Li, Yuxiao Nie+1

Department of Biostatistics, University of Michigan, Ann Arbor MI 48109

Brain-Computer Interfaces EEG Decoding

Open paper AskPDF arXiv

Electroencephalography (EEG)-based P300 brain-computer interfaces (BCIs) enable communication without physical movement by detecting stimulus-evoked neural responses. Accurate and efficient decoding remains challenging due to high dimensionality, temporal dependence, and complex interactions across EEG channels. Existing approaches often treat channels independently or rely on black-box models, limiting interpretability and personalization. We propose a sparse Bayesian time-varying regression framework that explicitly models pairwise EEG channel interactions while performing temporal feature selection, where a relaxed-thresholded Gaussian process prior induces structured sparsity in both channel-specific and interaction effects, enabling interpretable identification of task-relevant channels and channel pairs. Applied to a public P300 speller dataset of 55 participants, our method achieves a 96.4% median character-level accuracy using all sequence repetitions and attains the best overall decoding performance among competing statistical and deep learning approaches. Incorporating channel interactions yields subgroup-specific gains in character-level accuracy, particularly among participants abstained from alcohol (median 7%, up to 14%). Importantly, our approach improves median BCI-Utility by more than 10% at its optimal operating point, achieving peak throughput after only six sequence repetitions. These results demonstrate that explicitly modeling structured EEG channel interactions within a principled Bayesian framework enhances accuracy and user-centric throughput, and supports personalization in P300 BCI systems.

Abstract

Brain-Computer Interfaces

Explore similar work

Sparse Bayesian Modeling of EEG Channel Interactions Improves P300 Brain-Computer Interface Performance

When More Is Not Better: Component Anti-Synergy in a P300 Speller

Explainability of Recurrent Neural Networks for Enhancing P300-based Brain-Computer Interfaces