Diffusion weighted imaging is routinely used in the clinical settings. EPI is commonly used but suffers from severe geometric distortion. TSE/PROPELLER based DWI techniques are free of geometric distortion, but lead to long scan time and high SAR. In this project we propose a new SPLICE X-PROP technique to provide rapid high resolution DWI, by incorporating a SPLICE acquisition approach and a cross-blade GRASE readout into the PROPELLER technique. The improvement over conventional SPLICE PROPELLER technique is demonstrated with volunteer scans.
The new PROPELLER technique (hereinafter referred to as SPLICE X-PROP) incorporates a SPLICE3 (split-echo acquisition of FSE signal) scheme and a cross-blade GRASE4 (gradient- and spin-echo) readout. In TSE/PROPELLER DWI, the Carr-Purcell-Meiboom-Gill (CPMG) condition is violated. Compared to other solutions to the non-CPMG condition (such as the Alsop’s method1, X-Y phase modulation2, the LRX method5, etc), the SPLICE approach enables long echo train while preserving the full signal. SPLICE has been explored with PROPELLER, with the E1 and E2 echo groups placed in parallel (referred to as SPLICE PROPELLER, Fig. 1a)6 or orthogonal blade (PROPELLER DUO)7. The GRASE readout improves the acquisition efficiency and reduces the SAR compared to conventional TSE. There are various strategies to combine GRASE with PROPELLER, such as turbo-PROP8, X-PROP9, and steer-PROP10, which typically use the X-Y phase modulation2. The turbo-PROP method places the gradient and spin echoes in the same blade. It therefore requires advanced acquisition/reconstruction strategy to alleviate the artifacts due to eddy currents arising from the alternating readout gradients, and phase errors in the gradient echoes from field inhomogeneities and tissue susceptibilities. By separating the gradient and spin echoes into different blades, the X-PROP or steer-PROP method is less dependent on special reconstruction algorithms, and is therefore incorporated in the SPLICE X-PROP technique (Fig. 1b).
The sequence was implemented on a Philips 3T Ingenia scanner. Both parallel and orthogonal formats of the E1 and E2 groups, as well as the turbo- and X-PROP approaches were implemented, while only X-PROP with the parallel blade format was demonstrated in this abstract. All data were acquired with FOV = 240x240 mm2, resolution = 1x1 mm2, 20 slices with thickness = 5 mm and gap = 1.5 mm, one b = 0 and three b =1000 s/mm2 directions. SPLICE X-PROP data were acquired with ETL = 12, EPI factor = 3, refocusing flip angle = 90°, and scan time = 3:24. The reference conventional SPLICE PROPELLER data were acquired with ETL = 20, and scan time = 6:42 (without SENSE) and 3:16 (with SENSE reduction factor R = 2 to reduce scan time). With these parameters, SLICE X-PROP achieves a 38% reduction in SAR due to the refocusing RF pulse train, and a 80% faster speed compared to SPLICE PROPELLER (without acceleration). EPI data were also acquired to demonstrate the reduction of geometric distortion in the PROPELLER methods.
Fig. 2 compares the SPLICE X-PROP results with SPLICE PROPELLER and EPI. Both PROPELLER methods show no geometric distortion, which is commonly seen in EPI results (1st row). Without acceleration, SPLICE PROPELLER shows good image quality (2nd row). Parallel imaging can reduce the scan time to a clinically preferred level (~3 min) but leads to additional SNR loss due to the g-factor, as pointed to by the arrows in the b=1000 s/mm2 images and ADC map (3rd row). SPLICE X-PROP can achieve similar fast speed without the use of parallel imaging, and thus preserves the signals and exhibits decent image quality (4th row). The impact on effective b-factor from the different ETL/echo spacing was not accounted for11. Fig. 3 shows the full data set (isotropic b=1000 s/mm2 images and ADC maps) from a volunteer.