1355417 - View-Sharing and KWIC Filtering Deblurs 64-fold Accelerated Real-Time Phase-Contrast MRI Obtained with Radial k-space Sampling: A Preliminary Evaluation in Pediatric Congenital Heart Disease Patients

Real-time phase-contrast (rtPC) MRI has several advantages (faster, free-breathing, insensitive to arrhythmia) over clinical standard PC MRI. In pediatric patients with fast heart rates, it is particularly important to achieve high temporal resolution. We previously developed a 38.4-fold accelerated rtPC MRI pulse sequence with 41.7ms temporal resolution using radial k-space sampling and compressed sensing (CS), which produced 17% underestimation of peak velocity compared with clinical PC MRI¹. In this study, we sought to achieve higher temporal resolution (25ms; 64-fold accelerated) in rtPC MRI without significant blurring by exploiting view sharing (VS)² and k-space weighted image contrast (KWIC) filtering³.

Methods:

Data acquisition: We used existing raw k-space of 12 pediatric patients with congenital heart disease (8 males and 4 females, mean age = 11.0±3.2 yrs), who underwent a clinical PC and a rtPC MRI at the aortic valve, pulmonic valve, left and right pulmonary arteries. Imaging parameters are summarized in Figure 1a.

Image reconstruction: We compared three rtPC reconstruction methods: (i). Previously described CS with 5 radial spokes per frame (38.4-fold accelerated), (ii). CS with 3 radial spokes per frame (64-fold accelerated), (iii). CS with 3 native radial spokes per frame (64-fold accelerated) boosted with VS and KWIC filtering. Figure 1b illustrates the VS and KWIC schemes. CS was performed with temporal total variation¹.

Post-processing: We performed background phase correction⁴. The regions of interest (ROI) were manually delineated for rtPC and clinical PC. Same ROIs were used for the three rtPC methods.

Analysis: We computed the blur metric⁵ (0 [best] to 1 [worst]) to evaluate the image sharpness, which is important for ROI segmentation. The blur metrics were compared using ANOVA (Bonferroni correction). We measured the peak velocity, forward volume, backward volume, and regurgitant fraction for rtPC and clinical PC. We conducted linear regression to evaluate the rtPC methods.

Results:

Figure 2 shows representative images and flow measurements. CS with 3 radial spokes boosted by VS and KWIC filtering produced the best image quality and best flow curves with minimal underestimation of peak velocity. The mean blur metric was significantly (p < 0.001) better for CS 3 spokes with VS and KWIC (0.39±0.05) than CS 5 spokes (0.47±0.05) and CS 3 spokes (0.50±0.05). As a reference, the blur metric of clinical PC was 0.34±0.03.

Figure 3 shows the linear regression results. Hemodynamic results worsen from 5 to 3 spokes per frame, due to temporal blurring caused by over regularization in CS. However, adding VS and KWIC improves the results.

Conclusion:

This study demonstrates that VS and KWIC filtering deblurs 64-fold accelerated rtPC MR images. This reconstruction approach enables 25ms temporal resolution and relatively high accuracy in blood flow and velocity measurements. Future study includes more extensive testing in pediatric patients with congenital heart disease.