Arterial spin labeling (ASL) can measure quantitative perfusion noninvasively using blood as an endogenous tracer. Flow Alternating Inversion Recovery (FAIR) based ASL has been extensively studied for pulmonary perfusion imaging at 1.5T in research environments [1-4]. However, the clinical applications of FAIR have been limited largely due to the low signal to noise ratio (SNR), combined with bright signal from major pulmonary vasculature and potential image misregistration. The purpose of this study was three fold: 1) To evaluate FAIR at 3T for increased SNR due to higher field strength and prolonged T1 of blood and compare results against SPECT perfusion, 2) To combine FAIR with inflow saturation to suppress signal from major pulmonary vasculature, and 3) To combine FAIR with background suppression to minimize image misregistration artifacts. This was a prospective, HIPAA-compliant feasibility study. All subjects signed informed consent prior to imaging. The FAIR sequence was implemented on a 3T Ingenia scanner (Philips Healthcare, Best, The Netherlands). Similar to an earlier description [5], the sequence began with saturation of the imaging region, followed by labeling with a pair of selective and non-selective inversions using a hyperbolic secant pulse. A post-labeling delay of one cardiac cycle (i.e. R-R interval) [1] was used to allow labeled blood to perfuse the lungs. A Single-Shot Turbo Spin Echo (SShTSE) acquisition was used to minimize the susceptibility artifacts due to B₀ inhomogeneities in the lungs. The beginning of the sequence (i.e. saturation pulse) was ECG-triggered such that the data acquisition happens during the diastolic phase of the following cardiac cycle. Coronal and sagittal images were acquired with an in-plane resolution of 3x3 mm, 15mm slice thickness, TE of 44ms, and a TR of 3s using three pairs of control/label images in an 18-second breath hold. Quantification was performed using the perfusion equation for pulsed ASL [6] with the blood proton density taken as the average signal from an ROI drawn over the aorta in a separately acquired M₀ image. This sequence was evaluated in a cohort of 10 normal volunteers and compared against SPECT perfusion, which was performed in the same subjects during a 15-minute free breathing acquisition, following the injection of technetium-99m-macroaggregated albumin (Tc99m-MAA). Subsequently, the inflow saturation for the FAIR sequence was implemented using three equally-spaced saturation pulses applied over the labeling regions on either side of the imaging plane (fig. 1) during the final 500 ms prior to data acquisition. This inflow saturation suppresses the blood signal that flows into major pulmonary vasculature towards the end of the post-label delay. The FAIR sequence was also combined independently with a background suppression strategy using four non-selective inversion pulses applied during the post-label delay with optimal inversion times [7] (fig. 2).Figure 3 shows proton density, FAIR perfusion, quantified FAIR perfusion, and SPECT perfusion images in the coronal and sagittal planes in a representative volunteer. Perfusion was measured to be 508.9 $$$\pm$$$ 291.3 mL/100g/min (mean $$$\pm$$$ SD) across all subjects, consistent with previously reported values [8]. The FAIR perfusion images show details of the lungs with higher spatial resolution, but are also contaminated by the large pulmonary vessels, which are not present in the SPECT images (fig. 3, arrows). The FAIR perfusion images acquired with inflow saturation contain reduced signal in the major vessels, including aorta and larger pulmonary vessels, allowing for better evaluation of pulmonary parenchyma (fig. 4). Figure 5 shows the effect of background suppression to reduce image misregistration artifacts. Pulmonary perfusion appears similar between the two acquisition strategies, but background suppression significantly reduced the artifacts caused by breathing, seen in the diaphragm (red arrow) and fat (green arrow), and reduced background signal variations at the edges of the field of view (black arrow). Inflow saturation and background suppression improved the quality of FAIR perfusion-weighted images by reducing the signal in the major vasculature and minimizing image artifacts, respectively. In recent years, several statistical methods have been proposed to identify signal contribution from the major vasculature such that they can be eliminated from FAIR perfusion-weighted images for improved robustness in measuring pulmonary perfusion [3, 9]. The inflow saturation reported in this work eliminates the need for such processing, making the FAIR perfusion images equivalent in appearance to SPECT perfusion images. This inflow saturation, combined with the background suppression, should further improve FAIR perfusion images to make it a clinically feasible technique at 3T.