A Systematic Approach for Explaining Time and Frequency Features Extracted by Convolutional Neural Networks From Raw Electroencephalography Data

• Type of Material: Article
• Authors: Charles A Ellis, Emory University; Robyn L Miller, Emory University; Vince Calhoun, Emory University
• Language: English
• Date: 2022-05-31
• Publisher: FRONTIERS MEDIA SA
• Publication Version: Final Published Version
• Copyright Statement: © 2022 Ellis, Miller and Calhoun.
• License: Creative Commons Attribution 4.0 International
• Final Published Version (URL): http://dx.doi.org/10.3389/fninf.2022.872035
• Title of Journal or Parent Work: FRONTIERS IN NEUROINFORMATICS
• Volume: 16
• Start Page: 872035
• End Page: 872035
• Grant/Funding Information: This work was funded by the NIH grant R01MH123610 and NSF grant 2112455.
• Keywords: Neurosciences; explainable AI; Life Sciences & Biomedicine; CNNs; EEG; spectral explainability; Science & Technology; sleep stage classification; Neurosciences & Neurology; deep learning; SLEEP; Mathematical & Computational Biology
• Research Categories: Engineering, Biomedical; Computer Science

Abstract

In recent years, the use of convolutional neural networks (CNNs) for raw resting-state electroencephalography (EEG) analysis has grown increasingly common. However, relative to earlier machine learning and deep learning methods with manually extracted features, CNNs for raw EEG analysis present unique problems for explainability. As such, a growing group of methods have been developed that provide insight into the spectral features learned by CNNs. However, spectral power is not the only important form of information within EEG, and the capacity to understand the roles of specific multispectral waveforms identified by CNNs could be very helpful. In this study, we present a novel model visualization-based approach that adapts the traditional CNN architecture to increase interpretability and combines that inherent interpretability with a systematic evaluation of the model via a series of novel explainability methods. Our approach evaluates the importance of spectrally distinct first-layer clusters of filters before examining the contributions of identified waveforms and spectra to cluster importance. We evaluate our approach within the context of automated sleep stage classification and find that, for the most part, our explainability results are highly consistent with clinical guidelines. Our approach is the first to systematically evaluate both waveform and spectral feature importance in CNNs trained on resting-state EEG data.