SWI processing using a local phase difference modulated venous enhancement filter with noise compensation

Publication date: Available online 26 February 2019

Source: Magnetic Resonance Imaging

Author(s): Sreekanth Madhusoodhanan, Chandrasekharan Kesavadas, Joseph Suresh Paul

Abstract

Susceptibility weighted imaging (SWI) involves post-processing of gradient echo images which are sensitive to the spatial variations in magnetic susceptibility. The aim of this study is to develop an automated filtering scheme to enhance the contrast-to-noise ratio (CNR) and blooming on SWI. Here, the high-pass filtering for SWI processing is designed by applying a weighting function to the neighboring phase differences to enhance the susceptibility-related (SuR) contrast. This is accomplished by summing the neighboring phase differences, weighted with a scaled and shifted error function of the phase difference. Besides using the filter weights of this weighted high-pass (WHP) filter to minimize the filtering artefacts using a filter scale parameter, the CNR is further increased by introduction of the neighborhood-based noise compensation weights into the filtering process. These weights are deduced from the channel phase distribution, conditioned on the channel magnitude and noise variance. Using in vivo SWI data acquired at 1.5 T (16 nos.) and 3.0 T (30 nos.), the magnitude SWI processed using the noise compensated WHP (WHPC) filter is shown to provide an average CNR improvement of 68.40% over that of a homodyne high-pass (HHP) filter. Two tailed t-tests performed separately for different field strengths, show significant differences (p < 0.001) between mean separations of phase masks generated from the WHPC and HHP filtered phase images. In conclusion, the WHPC filter, tuned by the mean separation of the phase mask, enhances the SuR contrast of magnitude SWI for evaluation of mild cognitive impairments, brain tumor and hemorrhagic stroke.

https://ift.tt/2IDaPYy