Different Ventricular Fibrillation Types in Low-Dimensional Latent Spaces

Abstract

The causes of ventricular fibrillation (VF) are not yet elucidated, and it has been proposed that different mechanisms might exist. Moreover, conventional analysis methods do not seem to provide time or frequency domain features that allow for recognition of different VF patterns in electrode-recorded biopotentials. The present work aims to determine whether low-dimensional latent spaces could exhibit discriminative features for different mechanisms or conditions during VF episodes. For this purpose, manifold learning using autoencoder neural networks was analyzed based on surface ECG recordings. The recordings covered the onset of the VF episode as well as the next 6 min, and comprised an experimental database based on an animal model with five situations, including control, drug intervention (amiodarone, diltiazem, and flecainide), and autonomic nervous system blockade. The results show that latent spaces from unsupervised and supervised learning schemes yielded moderate though quite noticeable separability among the different types of VF according to their type or intervention. In particular, unsupervised schemes reached a multi-class classification accuracy of 66%, while supervised schemes improved the separability of the generated latent spaces, providing a classification accuracy of up to 74%. Thus, we conclude that manifold learning schemes can provide a valuable tool for studying different types of VF while working in low-dimensional latent spaces, as the machine-learning generated features exhibit separability among different VF types. This study confirms that latent variables are better VF descriptors than conventional time or domain features, making this technique useful in current VF research on elucidation of the underlying VF mechanisms.

Exploring Cardiac Rhythms and Improving ECG Beat Classification Through Latent Spaces

“…The decoder function f dec reconstructs the input data as approximated data x ′ from this latent space representation, h, using the weight matrix W dec and bias vector b dec [31]. The decoder function is denoted by:…”

“…Researchers have proposed various techniques to interpret DL models to uncover insights into their decisionmaking processes [28,29,30]. These include visualization methods such as Manifold Learning, an unsupervised set of techniques that give us visually interpretable representations and operates under the assumption that any observed data exist within a lower-dimensional manifold that is embedded in a higher-dimensional space [31]. These techniques identify low-dimensional structures within high-dimensional data [32].…”

“…Despite its potential, manifold learning techniques are still relatively unexplored in ECG analysis, with limited research papers dedicated to this area. For example, a study by Bernal et al [31] demonstrated the application of manifold learning in discriminating different types of ventricular fibrillation, showcasing its utility. A notable manifold learning technique is Uniform Manifold Approximation and Projection (UMAP) [33].…”

Volcanic Micro-Earthquake Classification With Spectral Manifolds in Low-Dimensional Latent Spaces

“…Psychoacoustic scales have been used to extract features for voice and audio analysis and for characterizing, analyzing, and modeling many other types of signals [22]- [24]. The psychoacoustic scales aim to extract information from audio signals regarding the perceptual characteristics of the environment and the listener [25].…”

Destabilization of the Organized Structure of Ventricular Fibrillation During Reperfusion

“…Провели частотно-амплитудный (спектральный) анализ электрограмм при перфузии, ишемии и реперфузии сердца при ФЖ методом быстрого преобразования Фурье в диапазоне частот 0,5-15 Гц, который является наилучшим методом для определения частотно-амплитудных параметров ФЖ [14].…”