Regresar

T-Wave Alternans Estimation with Manifold Learning

Abstract:

T-wave Alternans (TWA) measures the differences between consecutive T-waves in Electrocardiographic (ECG) studies, serving as an index for identifying high-risk sudden cardiac death patients. In this study, ECG Imaging (ECGI) is employed to assess the spatial distribution of alternans across the subject’s epicardium. ECGI signals were acquired from the torsos of eight patients with Long QT syndrome and three control subjects, while epicardial signals were estimated through the inverse problem of ECG. Conventional TWA estimation methods have limitations in harnessing the spatial and temporal richness of ECGI simultaneously, and to overcome this limitation, we propose using manifold learning (MnL). In this work, we compare linear and non-linear MnL approaches with conventional methods, evaluating the reconstruction error of MnL outputs, reaching minimum values of 0.55µV and 4.22µV with linear and non-linear methods, respectively. Our findings suggest that MnL methods can effectively leverage all the information provided by ECGI, facilitating the localization of areas that may exhibit higher levels of TWA.