Chapter 1: Invited Addresses and Tutorials on Signals, Coding,
  Systems and Intelligent Techniques

Title: Extracting a reference signal for cardiac MRI gating: Experimental study for wavelet functions choice

Author(s): Odette Fokapu, Dima Abi-Abdallah, André Briguet

Address: Laboratoire de Biomécanique et Génie Biomédical UMR6600 Université de Technologie de Compiègne 60205 Compiègne cedex, France | Laboratoire de Biomécanique et Génie Biomédical UMR6600 Université de Technologie de Compiègne 60205 Compiègne cedex, France | Laboratoire de RMN Méthodologie et Instrumentation en Biophysique UMR CNRS 5012 Université Claude Bernard Lyon 1 69616 Villeurbanne cedex, France

Reference: 12th International Workshop on Systems, Signals and Image Processing pp. 421-425

Abstract/Summary: An algorithm for extracting an efficient reference signal from an ECG contaminated by Nuclear Magnetic Resonance (NMR) artifacts, mainly for image synchronization and/or for subject monitoring, has been investigated. The proposed signal processing is based on filter bank decomposition using the wavelet transform, and was tested on various ECG signals recorded during three imaging sequences: Gradient Echo (GE), Fast Spin Echo (FSE) and Inversion Recovery with Spin Echo (IRSE). A significant part of this present work was devoted to the choice of wavelets and the number of scales necessary for the reference signal reconstruction. In order to determine the most appropriate wavelet for each excitation sequence a natural approach for wavelet selection is adopted. Based on bibliographical information and on an analysis of the noise generated by the imaging sequences, an optimal group of 14 wavelet basis functions, members of three wavelet families. (Daubechies, Coiflets, Symlets) was selected. The number of scales (8 levels) was fixed based on ECG spectral analysis, taking into account the frequency components of the contaminating artifacts. Only the sub-bands necessary for cardiac synchronization, those containing the essential part of the QRS energy were combined to form the reference signal. Using the set of selected wavelet functions, the efficiency of the presented method has been tested on a group of quite representative signals made up of highly contaminated simulated ECGs of normal and pathological heart beats (mean SNR <- 5dB), as well as some pathological rodents ECG records. For the all three sequences, the results have shown that an appropriate choice of the wavelet function could considerably improve the quality of the reference signal for a better MRI synchronization.

Order a copy of this article