Variability of perfusion dark rim artifacts due to Gibbs ringing
- 762 Downloads
KeywordsMyocardial Perfusion Imaging Perfusion Study Image Interpolation Bicubic Interpolation Zero Filling
Background and aims
Gibbs ringing is a well known source of Dark Rim Artifacts (DRA) in myocardial perfusion imaging . We examine the variability of this artifact. Specifically, we show that Gibbs artifacts are highly dependent on the edge position and that sub-pixel shifts can dramatically change the appearance.
Sub-pixel shifts were introduced in four in-vivo raw-data perfusion studies, where a DRA was visible. The shifts had a step of 1/8th of a pixel ranging from 0.125 to 0.875 of the in-plane pixel size. The unprocessed raw-data was phase-shifted using MATLAB before reconstructing it on the scanner using the same image processing as the original data.
The original perfusion study was done on a 1.5 T scanner (Avanto; Siemens): hybrid-EPI sequence with an EPI factor of 4; TR/TE of 5.1/1.7 ms; base resolution 128 pixels; pixel size 2.8 × 2.8 × 8 mm; flip angle 30°; bandwidth 1860 Hz pixel-1; TI (time of inversion) of 90 ms using a non-selective BIR-4 saturation pulse, TSENSE with R = 2. Perfusion was imaged during first pass of Gd-DTPA; stress induced by adenosine.
The original and shifted magnitude images were compared visually using CMRtools which applies sub-pixel interpolation in the image space.
The DRAs shown were carefully selected so as to not coincide with any known real perfusion defect.
Sub-pixel shifts in short-axis images were also simulated numerically in MATLAB. Images were also reconstructed with image based bicubic interpolation and zero filling with a factor of 4, and compared.
Results and discussion
Although the shift selected in Figure 1a–d was the worst-case, it should be recognised that this occurs by chance depending on image plane, in-plane offsets, and the patient's respiratory and cardiac motion. This is consistent with the random nature of DRA occurrence in clinical practice.
The visibility of Gibbs DRAs in perfusion studies is very dependent on the position of the subendocardial wall inside the pixel in the absence of zero-filled pre-FFT interpolation. Position variations from frame to frame in a typical gated perfusion study can explain some of the variability often seen in DRAs. Interpolation by zero-filling prior to enlarged FFT regularizes the DRA appearance.
This article is published under license to BioMed Central Ltd.