Ultrafast Intracranial Vessel Imaging With Non-Cartesian Spiral 3-Dimensional Time-of-Flight Magnetic Resonance Angiography at 1.5 T: An In Vitro and Clinical Study in Healthy Volunteers.

OBJECTIVES: Non-Cartesian spiral magnetic resonance (MR) acquisition may enable higher scan speeds, as the spiral traverses the k-space more efficiently per given time than in Cartesian trajectories. Spiral MR imaging can be implemented in time-of-flight (TOF) MR angiography (MRA) sequences. In this study, we tested the performance of five 3-dimensional TOF MRA sequences for intracranial vessel imaging at 1.5 T with qualitative and quantitative image quality metrics based on in vitro and in vivo measurements. Specifically, 3 novel spiral TOF MRA sequences (spiral-TOFs) and a compressed sensing (CS) technology-accelerated TOF MRA sequence (CS 3.5) were compared with a conventional (criterion standard) parallel imaging-accelerated TOF MRA sequence (SENSE).
MATERIALS AND METHODS: The SENSE sequence (5:08 minutes) was compared with the CS 3.5 sequence (3:06 minutes) and a spiral-TOF (spiral, 1:32 minutes), all with identical resolutions. In addition, 2 further isotropic spiral-TOFs (spiral 0.8, 2:12 minutes; spiral 0.6, 5:22 minutes) with higher resolution were compared with the SENSE. First, vessel tracking experiments were performed in vitro with a dedicated vascular phantom to determine possible differences in the depiction of cross-sectional areas of vessel segments. For the in vitro tests, an additional 3-dimensional proton density-weighted sequence was added for comparison reasons. Second, 3 readers blinded to sequence details assessed qualitative (16 features) and 2 readers assessed quantitative (contrast-to-noise ratio [CNR], contrast ratio [CR], vessel sharpness, and full width at half maximum edge criterion measurements) image quality based on images acquired from scanning 10 healthy volunteers with all 5 TOF sequences. Scores from quantitative image quality analysis were compared with Kruskal-Wallis, analysis of variance, or Welch's analysis of variance, followed by Dunnett's or Dunnett's T3 post hoc tests. Scores from qualitative image quality analysis were compared with exact binomial tests, and the level of interreader agreement was determined with Krippendorff's alpha.
RESULTS: Concerning the in vitro tests, there were no significant differences between the 5 TOFs and the proton density-weighted sequence in measuring cross-sectional areas of vessel segments (P = 0.904). As for the in vivo tests, the CS 3.5 exhibited equal qualitative image quality as the SENSE, whereas the 3 spiral-TOFs outperformed the SENSE in several categories (P values from 0.002 to 0.031). Specifically, the spiral 0.8 and 0.6 sequences achieved significantly higher scores in 12 categories. Interreader agreement ranged from poor (alpha = -0.013, visualization of internal carotid artery segment C7) to substantial (alpha = 0.737, number of vessels visible, sagittal). As for the quantitative metrics, the CS 3.5 and all 3 spiral-TOFs presented with significantly worse CNR than the SENSE ([mean ± SD] SENSE 37.48 ± 7.13 vs CS 3.5 31.14 ± 5.97 vs spiral 19.77 ± 1.65 vs spiral 0.8 16.18 ± 2.14 vs spiral 0.6 10.37 ± 1.05). The CR values did not differ significantly between the SENSE and the other TOFs except for the spiral sequence that showed significantly improved CR (SENSE 0.53 ± 0.03 vs spiral 0.56 ± 0.03). As for vessel sharpness, the SENSE was outperformed by all spiral-TOFs (SENSE 0.37 ± 0.03 vs spiral 0.52 ± 0.07 vs spiral 0.8 0.53 ± 0.08 vs spiral 0.6 0.73 ± 0.09), whereas the CS 3.5 performed equally well (SENSE 0.37 ± 0.03 vs CS 3.5 0.37 ± 0.03). Full width at half maximum values did not differ significantly between any TOF.
CONCLUSIONS: Spiral-TOFs may deliver high-quality intracranial vessel imaging thus matching the performance of conventional parallel imaging-accelerated TOFs (such as the SENSE). Specifically, imaging can be performed at unprecedented scan times as short as 1:32 minutes per sequence (70.12% scan time reduction compared with SENSE). Optionally, spiral imaging may also be used to increase spatial resolution while maintaining the scan time of a Cartesian-based acquisition schema. The CNR was decreased in spiral-TOF images.