Eliminating dark-rim artifacts in first-pass myocardial perfusion imaging

Behzad Sharif,Chrisandra Shufelt,Troy Labounty,Noel Bairey Merz,Rohan Dharmakumar,Daniel S Berman,Louise E Thomson,Debiao Li

doi:10.1186/1532-429x-15-s1-o3

Behzad Sharif, Chrisandra Shufelt + Show 6 more

Open Access

https://doi.org/10.1186/1532-429x-15-s1-o3

Copy DOI

Abstract

Background We demonstrate that projection imaging significantly reduces the prevalence and spatial extent of subendocardial dark-rim artifacts (DRAs) in first-pass perfusion (FPP) myocardial MR, compared to conventional Cartesian techniques. A major cause of DRAs, which remain a major concern in FPP imaging, is known to be the socalled Gibbs ringing (truncation) phenomenon [1-3]. Radial k-space sampling exhibits minimal Gibbs effects with typical FPP parameters, thereby eliminating a major contributing factor to DRAs [4]. The underlying theory is demonstrated in Fig. 1, which describes Cartesian and radial k-space sampling (with the same number of readouts) and the corresponding point spread functions (PSFs). Insufficient coverage along phase-encode direction with Cartesian sampling results in significant ringing in image domain (Fig. 1b). In contrast, angular undersampling results in streaks outside of a “local” region for

Highlights

We demonstrate that projection imaging significantly reduces the prevalence and spatial extent of subendocardial dark-rim artifacts (DRAs) in first-pass perfusion (FPP) myocardial MR, compared to conventional Cartesian techniques
Two FPP scans (SR-prepared FLASH) were performed at rest (>10 minutes gap) using a single-shot radial pulse sequence followed by a singleshot Cartesian sequence
Representative images from 4 of the 12 studied subjects are shown in Fig. 2, where the top panels show Cartesian images and bottom ones are the corresponding radial images

Summary

Background

We demonstrate that projection imaging significantly reduces the prevalence and spatial extent of subendocardial dark-rim artifacts (DRAs) in first-pass perfusion (FPP) myocardial MR, compared to conventional Cartesian techniques. The underlying theory is demonstrated, which describes Cartesian and radial k-space sampling (with the same number of readouts) and the corresponding point spread functions (PSFs). Insufficient coverage along phase-encode direction with Cartesian sampling results in significant ringing in image domain (Fig. 1b). Insufficient k-space coverage along Ky (phase-encode direction) results in significant ringing along y, as shown in (b). Panels (d)-(f) show reconstructions of an MR gelatin-Gadolinium phantom with realistic signal intensity ratios, demonstrating robustness of projection imaging to Gibbs ringing: (d) fully sampled (ground truth) image with 1x1 mm resolution (384x384 matrix); (e) Cartesian imaging with 108 phase-encodes (arrows point to DRA); (f) radial imaging with 108 projections (no DRAs, mild streaking). Panels 1d-f show phantom studies (Gelatin-based with realistic contrast ratios, resembling the LV with a deficit region) verifying the described PSF effects

Methods

Results

Conclusions