Learning Exemplar Representations in Single-Trial EEG Category Decoding

TL;DR

This paper shows that EEG-based object category decoding models can learn object representations even with simple training, raising concerns about potential data leakage inflating reported accuracies in previous studies.

Contribution

It demonstrates that common EEG classification methods can inadvertently learn object representations due to trial overlap, questioning the validity of prior high-accuracy results.

Findings

Simple classifiers can learn object representations from EEG data with category labels.

Potential data leakage may have inflated the accuracy of previous EEG decoding studies.

Results suggest the need for careful experimental design to avoid overlap-related biases.

Abstract

Within neuroimgaing studies it is a common practice to perform repetitions of trials in an experiment when working with a noisy class of data acquisition system, such as electroencephalography (EEG) or magnetoencephalography (MEG). While this approach can be useful in some experimental designs, it presents significant limitations for certain types of analyses, such as identifying the category of an object observed by a subject. In this study we demonstrate that when trials relating to a single object are allowed to appear in both the training and testing sets, almost any classification algorithm is capable of learning the representation of an object given only category labels. This ability to learn object representations is of particular significance as it suggests that the results of several published studies which predict the category of observed objects from EEG signals may be affected by a subtle form of leakage which has inflated their reported accuracies. We demonstrate the ability of both simple classification algorithms, and sophisticated deep learning models, to learn object representations given only category labels. We do this using two datasets; the Kaneshiro et al. (2015) dataset and the Gifford et al. (2022) dataset. Our results raise doubts about the true generalizability of several published models and suggests that the reported performance of these models may be significantly inflated.