Faculdade de Ciências e Tecnologia (FCT) >
FCT Departamentos >
FCT: Departamento de Física >
FCT: DF - MA Dissertations >
Please use this identifier to cite or link to this item:
|Title: ||Correction of spatial distortion in magnetic resonance imaging|
|Authors: ||Mendes, Pedro Mota|
|Advisor: ||Secca, Mário|
|Keywords: ||Image quality|
Magnetic resonance imaging (MRI)
Spatial distortion correction
|Issue Date: ||2011|
|Publisher: ||Faculdade de Ciências e Tecnologia|
|Abstract: ||Magnetic Resonance Imaging (MRI) has been a major investigation and research focus among
scientific and medical communities. So, new hardware with superior magnetic fields and faster sequences has been developed. However, these improvements result in intensity and spatial distortions, particularly in fast sequences, as Echo Plana Imaging (EPI), used in functional and
diffusion-weighed MRI (fMRI and DW-MRI). Therefore, correction of spatial distortion is useful to obtain a higher quality in this kind of images.
This project contains two major parts. The first part consists in simulating MRI data required for assessing the performance of Registration methods and optimizing parameters. To assess the methods five evaluation metrics were calculated between the corrected data and an undistorted EPI, namely: Root Mean Square (RMS); Normalized Mutual Information (NMI), Squared Correlation
Coefficient(SCC); Euclidean Distance of Centres of Mass (CM) and Dice Coefficient of
segmented images. In brief, this part validates the applied Registration correction method. The project’s second part includes correction of real images, obtained at a Clinical Partner. Real images are diffusion weighted MRI data with different b-values (gradient strength coefficient), allowing performance assessment of different methods on images with increasing b-values and decreasing SNR. The methods tested on real data were Registration, Field Map correction and a new proposed pipeline, which consists in performing a Field Map correction after a registration process.
To assess the accuracy of these methods on real data, we used the same evaluation metrics, as for simulated data, except RMS and Dice Coefficient.
At the end, it was concluded that Registration-based methods are better than Field Map, and that the new proposed pipeline produces some improvements in the registration. Regarding the influence of b-value on the correction, it is important to say that the methods performed using images with higher b’s showed more improvements in regarding metric values, but the behaviour is similar for all b-values.|
|Description: ||Dissertation to Obtain the Degree of Master
in Biomedical Engineering|
|Appears in Collections:||FCT: DF - MA Dissertations|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.