Data di Pubblicazione:
2017
Citazione:
ADAPTIVE MODELS-BASED CARDIAC SIGNALS ANALYSIS AND FEATURE EXTRACTION / A. Kheirati Roonizi ; advisor: R. Sassi ; director of doctoral school: P. Boldi. DIPARTIMENTO DI INFORMATICA, 2017 Feb 28. 29. ciclo, Anno Accademico 2016. [10.13130/kheirati-roonizi-ebadollah_phd2017-02-28].
Abstract:
Signal modeling and feature extraction are among the most crucial and important
steps for stochastic signal processing. In this thesis, a general framework that employs
adaptive model-based recursive Bayesian state estimation for signal processing and
feature extraction is described. As a case study, the proposed framework is studied
for the problem of cardiac signal analysis. The main objective is to improve the signal
processing aspects of cardiac signals by developing new techniques based on adaptive
modelling of electrocardiogram (ECG) wave-forms. Specially several novel and
improved approaches to model-based ECG decomposition, waveform characterization
and feature extraction are proposed and studied in detail. In the concept of ECG decomposition
and wave-forms characterization, the main idea is to extend and improve
the signal dynamical models (i.e. reducing the non-linearity of the state model with
respect to previous solutions) while combining with Kalman smoother to increase the
accuracy of the model in order to split the ECG signal into its waveform components,
as it is proved that Kalman filter/smoother is an optimal estimator in minimum mean
square error (MMSE) for linear dynamical systems. The framework is used for many
real applications, such as: ECG components extraction, ST segment analysis (estimation
of a possible marker of ventricular repolarization known as T/QRS ratio) and
T-wave Alternans (TWA) detection, and its extension to many other applications is
straightforward.
Based on the proposed framework, a novel model to characterization of Atrial Fibrillation
(AF) is presented which is more effective when compared with other methods
proposed with the same aims. In this model, ventricular activity (VA) is represented
by a sum of Gaussian kernels, while a sinusoidal model is employed for atrial activity
(AA). This new model is able to track AA, VA and fibrillatory frequency simultaneously
against other methods which try to analyze the atrial fibrillatory waves (f-waves)
after VA cancellation.
Furthermore we study a new ECG processing method for assessing the spatial dispersion
of ventricular repolarization (SHVR) using V-index and a novel algorithm to
estimate the index is presented, leading to more accurate estimates. The proposed algorithm
was used to study the diagnostic and prognostic value of the V-index in patients
with symptoms suggestive of Acute Myocardial Infraction (AMI).
Tipologia IRIS:
Tesi di dottorato
Elenco autori:
A. Kheirati Roonizi
Link alla scheda completa:
Link al Full Text: