Improved compressed sensing reconstruction for [Formula: see text]F magnetic resonance imaging
In magnetic resonance imaging (MRI), compressed sensing (CS) enables the reconstruction of undersampled sparse data sets. Thus, partial acquisition of the underlying k-space data is sufficient, which significantly reduces measurement time. While F MRI data sets are spatially sparse, they often suffe...
Saved in:
Published in | Magma (New York, N.Y.) Vol. 32; no. 1; pp. 63 - 77 |
---|---|
Main Authors | , , , , , , , , , , , , |
Format | Journal Article |
Language | English |
Published |
Germany
01.02.2019
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | In magnetic resonance imaging (MRI), compressed sensing (CS) enables the reconstruction of undersampled sparse data sets. Thus, partial acquisition of the underlying k-space data is sufficient, which significantly reduces measurement time. While
F MRI data sets are spatially sparse, they often suffer from low SNR. This can lead to artifacts in CS reconstructions that reduce the image quality. We present a method to improve the image quality of undersampled, reconstructed CS data sets.
Two resampling strategies in combination with CS reconstructions are presented. Numerical simulations are performed for low-SNR spatially sparse data obtained from
F chemical-shift imaging measurements. Different parameter settings for undersampling factors and SNR values are tested and the error is quantified in terms of the root-mean-square error.
An improvement in overall image quality compared to conventional CS reconstructions was observed for both strategies. Specifically spike artifacts in the background were suppressed, while the changes in signal pixels remained small.
The proposed methods improve the quality of CS reconstructions. Furthermore, because resampling is applied during post-processing, no additional measurement time is required. This allows easy incorporation into existing protocols and application to already measured data. |
---|---|
Bibliography: | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23 |
ISSN: | 1352-8661 |
DOI: | 10.1007/s10334-018-0729-1 |