Autoencoder-Based Models for Scalp EEG: A Systematic Review of Architectures, Applications, Latent Representations, Interpretability, and Validation

Abstract

This registration sets out the protocol for a systematic review and evidence map of peer-reviewed empirical studies using autoencoder-based models with recorded non-invasive scalp EEG. The review will examine architectures, EEG representations, application areas, latent-space designs, interpretation methods, validation practices, methodological quality, and reproducibility. Searches will cover Scopus, Web of Science Core Collection, PubMed, and IEEE Xplore. Springer Nature Link will serve as a supplementary publisher-platform search. Backward and forward citation searching will supplement the electronic searches. The final reproducible searches, formal screening, extraction, methodological appraisal, and synthesis had not started when this registration was submitted.