|Nicole Brueschweiler||Master project|
|Section Microtechnique, EPFL||April 2007|
This project was done in collaboration with the Institute for Biomedical Engineering at ETHZ.
The task of this master project was to apply different wavelet-based denoising algorithms on magnetic resonance signals. Besides the already well known wavelet shrinkage method, also a recently developed SURE-algorithm was implemented. The SURE-algorithm uses the undecimated wavelet transform and a parametric, pointwise thresholding function whose parameters were estimated using a global SURE optimum. The denoising was done on the complex FID and the real spectrum of the MRS signals using standard wavelets and the newly developed Exponential Spline Wavelet. The denoising methods were compared using the signal-to-noise ratio and two performance measures which are specific to MRS.
A further task of the project was the creation of a graphical user interface summarizing the different denoising methods and different wavelets.
Fig. 1: Noisy MR-Spectrum (in blue) and denoised spectrum (in red). The denoising was done using the SURE-algorithm and the Haar wavelet.
Fig. 2: Part of the user interface for the zero- and first-order phase correction.