Stroke is a leading cause of mortality and morbidity worldwide and arises from diverse intracranial wall pathologies. Accurate characterization of these pathologies may help elucidate stroke etiology and facilitate treatment decision. High-resolution MR using variable-flip-angle 3D fast spin-echo (FSE) has emerged as a promising intracranial vessel wall imaging technique [1-4]. However, its typical implementations on clinically available MR systems (e.g. 3T) have several limitations such as suboptimal image contrast weighting, inadequate cerebrospinal fluid (CSF) signal attenuation, small spatial coverage, and long scan times. This work aimed to develop a 3D FSE-based method that allows for CSF-attenuated T1-weighted whole-brain vessel wall imaging within 8 min. Theory The technical development was conducted in two phases. First, a commercially available 3D FSE sequence (SPACE, Siemens) was optimized in spatial coverage and contrast weighting. A non-selective hard RF pulse is used for excitation to achieve whole-brain coverage. This allows the TE to reduce, thus increasing both T1-weighting and SNR. To further suppress CSF signals and enhance vessel wall delineation, a flip-down RF pulse module is applied immediately after each refocusing pulse train, in effect serving as an inversion-recovery (IR) preparation [5]. Second, the imaging time based on the above optimized IR-SPACE sequence was further reduced. Elliptical data sampling and prolonged echo train length (ETL) were exploited to expedite the acquisition. However, this would reduce overall SNR. On the other hand, SNR is intimately related to the refocusing flip angles that are calculated for a prescribed signal evolution of a tissue with specific T1 and T2 values (denoted here as simulation T1 and T2) [6]. Thus, SNR may be gained by using appropriate simulation T1/T2 values. Experiment Imaging was conducted on a 3T system with a 32-channel head coil. Phase I. IR-SPACE was validated on volunteers (7 health, 4 patients with severe luminal stenosis) in comparison with the conventional SPACE sequence. The major imaging parameters shared by the two sequences were: 3D sagittal orientation; TR/TE 800/10ms; isotropic spatial resolution 0.5mm; GRAPPA factor 2; ETL 37; signal average 1; scan time 11-12min. Moreover, vendor’s default simulation T1/T2 (940/100ms) were used in all scans. Phase II. the effects of simulation T2 (T1 has less effect on SNR according to pilot data and was fixed at 1100ms) and ETL on vessel wall SNR, wall-CSF CNR, and white-gray matter CNR (indicative of T1 contrast weighting) were first explored on 11 healthy subjects. The range of potentially useful protocols (i.e. combinations of ETL and simulation T2) was then narrowed; specifically, ETL=52 combined with a simulation T2 of 140, 170, and 200 ms were respectively tested on additional 10 healthy subjects. A combination of ETL=36/T2=100ms as used in Phase I was used as the reference. An optimal imaging protocol was determined from the four scans and finally applied to a study of 18 patients for preliminary validation. Phase I The new sequence provided spatial coverage for the entire intracranial arterial system. Smaller vessel segments such as the middle cerebral artery M3 and M4 were visible. The contrast ratio of vessel wall-to-CSF (0.14±0.16 vs. 0.52±0.30, p = 0.007) and vessel wall sharpness (0.89±0.61mm-1 vs. 1.22±0.56mm-1, p < 0.001) were significantly enhanced as compared with the conventional sequence. The boosted T1 contrast weighting and CSF attenuation made the sequence more sensitive to high-T1-signal features (Fig. 1). Both atherosclerosis and inflammatory wall disease were identified by IR-SPACE. Phase II Increasing simulation T2 improved SNR and CNR (Fig. 2 a, b), and ETL had an opposite effect (Fig. 2 c, d). An ETL of 52 appeared to allow the scan time to reduce to 8 min while avoiding drastic SNR/CNR sacrifice. The use of ETL=52/T2=170ms significantly increased wall SNR (p=0.002), wall-CSF CNR (0.026), and white-gray matter CNR (0.004), but reduction of vessel wall sharpness (inner and outer boundary) was not significant, as compared with the original 12-min protocol (ETL=36/T2=100ms) (Fig. 3). This combination was chosen as an optimal protocol with which wall abnormalities were detected in all patients in agreement with their clinical diagnosis (Fig. 4). The 8-min scan was well tolerated in all patients according to a post-scan survey and image quality evaluation.Whole-brain intracranial vessel wall evaluation is feasible with clinically acceptable scan time and quality. Our preliminary data has shown that the proposed method offers good vessel wall delineation and accurate identification of vessel wall abnormalities due to improved T1 contrast weighting and CSF attenuation, large spatial coverage, and acceptable scan times. A large-scale trial on using the technique for diagnosis of stroke etiology is underway to establish its clinical usefulness.