Effects of k-space filtering and image interpolation on image fidelity in 1H MRSI

Abstract

2D MRSI suffers from the effect of the spatial response function due to the truncation of the sampling of k-space. Filtering of the k-space data-set is often used to suppress the side lobes caused by the effects of the SRF, where the sampled data-set is multiplied with a weighting function before inverse FT. Commonly used filters in MRSI are the cosine, Hanning and Hamming filters. The data-set is often interpolated into a larger image matrix size for analysis, where “Fourier interpolation” (FoI) and “cubic spline interpolation” (CSpI) are two common methods. In this work, the effects of k-space filtering in practical usage was examined, and the image representations of the object for the two interpolation methods were compared. This study showed that application of filtering improves the image representation of the structures in the object and should be used in MRSI. FoI correctly visualizes the information inherent in the data-set, while the features of the object were dependent on the position of the object relative the original matrix in the CSpI interpolated images. FoI should therefore be used for quantitative evaluation of MRSI images.