First-pass perfusion CMR with reduced dark-rim artifact and instantaneous image reconstruction using optimized cartesian sampling and apodization

Background Subendocardial dark-rim artifact (DRA) continues to be a major issue that limits the diagnostic performance of first pass perfusion (FPP) CMR. Non-Cartesian approaches such as radial or spiral acquisition have been proposed to minimize DRAs. Among these approaches, those that can operate without the need for breath-holding typically require a time-consuming offline image reconstruction procedure, which limits their clinical accessibility. We propose a free-breathing DRA-reduced FPP scheme with instant image reconstruction on the scanner.


Background
Subendocardial dark-rim artifact (DRA) continues to be a major issue that limits the diagnostic performance of first pass perfusion (FPP) CMR. Non-Cartesian approaches such as radial or spiral acquisition have been proposed to minimize DRAs. Among these approaches, those that can operate without the need for breath-holding typically require a time-consuming offline image reconstruction procedure, which limits their clinical accessibility. We propose a free-breathing DRA-reduced FPP scheme with instant image reconstruction on the scanner.

Methods
Healthy subjects (n=10) were studied on a 3T Siemens scanner using the standard 12-channel coil. Two freebreathing FPP scans (3 slice coverage) were performed at rest, first using the proposed method followed by a conventional method (15 minute gap).
Apodization of k-space data is known to reduce Gibbs ringing effects, a major underlying cause of the DRA. It also enhances signal-to-noise (SNR) but comes at the cost of lower resolution. In the proposed method, a high parallel imaging factor (4 fold) with Cartesian sampling is used to acquire a high-resolution FPP scan (in-plane: 1.7 x 1.7 mm 2 ). Following parallel imaging reconstruction, an optimized apodizer is applied, which both reduces Gibbs ringing and improves SNR. The apodization level was adjusted so that the apodized images had the same resolution as the conventional FPP scan (in-plane: 2.7 x 2.2 mm 2 ).
Apodization was implemented as an additional postfiltering step and it was integrated to the inline image reconstruction routine of the scanner. Images were scored by 2 expert readers blinded to the protocol (consensus artifact scoring; scale: 0 = No DRA, 4 = Severe DRA). In addition, dogs with myocardial infarction (n=2) and a control dog were studied using a similar setting. Fig. 1(a,b) shows representative images and the histograms of DRA scores are presented in Fig. 1(c,d). The mean DRA score for the proposed method is lower than the conventional method (p<0.01). More importantly, the number of slices with severe DRA (scores 3 or 4) is significantly lower for the proposed method. Image quality scores were also assigned by the readers (0-4 scale; 0: poor; 4: excellent). The difference of image quality between the two methods was not significant (3.5 for the proposed method and 3.9 for conventional; p=0.2). Fig. 2 provides results from the canine studies, demonstrating that FPP images acquired with proposed method can detect small subendocardial defects corresponding to infarct territories (LGE images).

Conclusions
We developed an optimized Cartesian sampling and apodization scheme for free-breathing perfusion CMR. Compared to the conventional FPP imaging method, the severity and prevalence of subendocardial DRAs were significantly reduced. A key feature of the proposed method is that it uses the widely-available parallel imaging reconstruction software to achieve near-instant reconstruction on the scanner, potentially enhancing its accessibility in clinical studies.   (a1,b1): images correspond to a chronic MI dog study that shows small subendocardial LGE (a) and FPP defect in the lateral wall. The defect is clearly seen demonstrating the ability of the proposed method to resolve small defects (following apodization). (a2,b2) Similar result with a larger defect region in an acute MI dog. (a3,b3): images for a control animal study. All images correspond to the peak myocardial enhancement phase. Among the 3 studies, minimal DRA is seen in only (b2) and (b1,b3) are free of DRA. The results demonstrate that reduction of subendocardial DRAs using apodization (applied to a high-resolution scan) does not lead to elimination/masking of subendocardial defects in the reconstructed apodized images.