Canine preictal topology: ordinal complexity and neural mapping

Abstract

This research explores the identification of preictal biomarkers in canine epilepsy by employing a multiscale analysis of intracranial EEG data. The approach integrates entropy and complexity quantification using the Bandt–Pompe method plane) with topological feature extraction via Self-Organizing Maps (SOMs) and Uniform Manifold Approximation and Projection (UMAP). Although the entropy-complexity framework captured subject-specific neural characteristics, it did not succeed in distinguishing between preictal and interictal states. In contrast, the SOM-UMAP pipeline revealed clear preictal markers, attributed to the reconfiguration of the mesoscale network using optimal parameters The main contributions of this study include the topological differentiation of brain states beyond the reach of traditional methods, the discovery of individualized epileptogenic patterns in UMAP embeddings, and the development of a validated methodology suitable for implantable device applications. By combining ordinal pattern analysis with topological preservation techniques, this work advances both the theoretical understanding of seizure mechanisms and the practical implementation of personalized seizure prediction tools, outperforming conventional univariate strategies to detect latent preictal signatures.