The paper introduces an Extended Kalman Smoother (EKS) for T-wave alternans (TWA) detection, based on a dynamical model which is not directly dependent on amplitude (EKS3). In this framework we consider separate states for PQRS and an amplitude-free state model for T-wave. There are some theoretical advantages that EKS3 has over other frameworks recently introduced with the same aims (e.g., EKS6-4obs, Akhbari et al., 2014 and EKS25-4obs, Akhbari et al., 2013). For instance, no longer depending directly on the amplitude of the Gaussian kernel, it is able to model the nuances in the T-waves, even when small or abrupt changes happen in the signal. Moreover, it reduces the nonlinearity of the model and it uses only three states, resulting in a significant decrease in complexity. We compared the proposed method with EKS6-4obs and EKS25-4obs using data from the 2008 Physionet TWA challenge dataset. While all the methods showed similar performances in estimating the average TWA value, the reduced standard deviation displayed by EKS3 facilitates the adjudication of TWA's presence, when it assumes small values.
A signal decomposition based Kalman smoother for T-wave alternans detection / A. Kheirati Roonizi, R. Sassi - In: 2015 AEIT International Annual Conference (AEIT)[s.l] : IEEE Press, 2015. - ISBN 9788887237283. - pp. 1-4 (( convegno AEIT International Annual Conference tenutosi a Napoli nel 2015 [10.1109/AEIT.2015.7415278].
A signal decomposition based Kalman smoother for T-wave alternans detection
A. Kheirati Roonizi;R. Sassi
2015
Abstract
The paper introduces an Extended Kalman Smoother (EKS) for T-wave alternans (TWA) detection, based on a dynamical model which is not directly dependent on amplitude (EKS3). In this framework we consider separate states for PQRS and an amplitude-free state model for T-wave. There are some theoretical advantages that EKS3 has over other frameworks recently introduced with the same aims (e.g., EKS6-4obs, Akhbari et al., 2014 and EKS25-4obs, Akhbari et al., 2013). For instance, no longer depending directly on the amplitude of the Gaussian kernel, it is able to model the nuances in the T-waves, even when small or abrupt changes happen in the signal. Moreover, it reduces the nonlinearity of the model and it uses only three states, resulting in a significant decrease in complexity. We compared the proposed method with EKS6-4obs and EKS25-4obs using data from the 2008 Physionet TWA challenge dataset. While all the methods showed similar performances in estimating the average TWA value, the reduced standard deviation displayed by EKS3 facilitates the adjudication of TWA's presence, when it assumes small values.File | Dimensione | Formato | |
---|---|---|---|
07415278.pdf
accesso riservato
Tipologia:
Publisher's version/PDF
Dimensione
588.79 kB
Formato
Adobe PDF
|
588.79 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.