3He MR imaging was performed at 3.0T (Signa EXCITE, GE Healthcare, Waukesha, WI) using a transmit-receive bird cage coil (Morris Instruments, Ottawa, Canada), with a length of 17.15 cm and diameter of 11.8 cm, tuned to the 3He resonance (97.3 MHz) and a high performance insert gradient coil (G = 17 G/cm, slew rate = 1500 mT/m/s) described previously (12). 3He gas was polarized using a spin-exchange optical pumping system (HeliSpin; GE Healthcare, Durham, NC). Optical pumping of the gas was typically performed for 18 hr, resulting in polarizations in excess of 32%. The hyperpolarized 3He gas was transferred to a Tedlar bag previously rinsed three times with medical-grade N2 gas and subsequently vacuumed (100 mtorr) to minimize depolarization of gas caused by 3He interactions with O2. The bag was then placed in the pressurized reservoir attached to the ventilator in the bore of the MR imaging magnet 60 cm from the isocenter. The T1 of the gas in the reservoir within the MR imaging magnet was measured to be ≈43 min.
Single-slice FAVOR images were obtained in the coronal plane using a fast gradient-recalled echo pulse sequence (TE = 1 ms, TR = 3.3 ms, FOV = 4 cm, Nx = 128, Ny = 128) using reverse centric k-space sampling and incorporating VFA pulses triggered by the ventilator following each 3He breath (8 breaths, 2 sec apart) with no air wash-out breaths between 3He breaths. Reverse centric sampling was chosen to allow more time for 3He to diffuse to the periphery of the lungs before acquiring the center of k-space. Each image acquisition was triggered by the ventilator at peak inspiration and required ≈430 ms, during which time respiratory motion was suspended within the 2-sec breathing cycle. For each breath (i.e., image), n, the VFA RF pulse trajectory was calculated using the following equation (13):
where the number of the phase encode step i = 1…128 and the number of RF pulses applied Nn = 1024, 896, 768, …, 128 for breaths n = 1, 2, 3, …, 8, respectively. T1,n = 31, 42, 53, 73, 82, 87, 88.5, 90 sec represent the longitudinal relaxation times due to O2 in the lung previously measured during breath-holds on a representative rat after n successive wash-in 3He breaths (14). However, the speed of the FAVOR technique with respect to T1,O2 means this correction is small (this will be discussed later). The RF pulses were calibrated by adjusting the transmitter gain until no measurable change in signal over 128 pulses over the entire sample (i.e., without phase encoding) was obtained following a single 3He breath. If too much power was applied per RF pulse the received signal from successive pulses decreased; and conversely, if too little power was applied the received signal gradually increased and the response from the last pulse, which had α = 90°, was large (Fig. 2). The slight upward trend at the tail of the uniform response (squares in Fig. 2) is due to a B1 inhomogeneity of 2.47% over a 4 × 4 × 4 cm volume. It was calculated as the coefficient of variation of the flip angle using a balloon phantom with a long T1 compared to the acquisition time such that B1 could be measured. Simulations show that 2% inhomogeneity gives a mean full-width at half-maximum of the point spread function of 1.23 compared to the ideal value of 1.21 for the case of no inhomogeneity. Five to eight single breaths of 3He were usually required for VFA calibration. Figure 3 shows the VFA RF pulse trajectory for an eight-image acquisition. As a further calibration check, the number of pulses was extended to 1024 following a single 3He breath and no significant change in signal was observed, confirming that signal loss due to RF depolarization and T1 relaxation was effectively offset by the FAVOR technique. Two preparatory “dummy” breaths of 3He were supplied by the ventilator prior to data acquisition to bleed the portion of the 3He supply line inside the RF coil containing 3He, which was depolarized during the course of the previous scan.
For image analysis a lower threshold was set by visual inspection of the last image from a series of breaths, such that all the voxels not adjacent to the lung (i.e., background) were removed. The remaining voxels were used to create a mask which was then applied to all images in the series. Ventilation images were also obtained in the conventional way, i.e., following successively increasing numbers of 3He wash-in breaths (up to eight breaths, tidal volume ≈2.6 mL, PIP = 12 cm H2O) separated by intervals of 80 air wash-out breaths (1). However, a 128 pulse VFA sequence was used to image the last breath in each cycle to avoid any blurring due to nonuniform signal amplitudes from using a constant flip angle sequence. Since the pressurized reservoir used here is known to cause variations in tidal volumes over the large numbers of breaths required by the conventional technique (>50), due to collapse of the Tedlar bag despite the active pressurizing with a pump, a full bag was used for each conventional measurement, resulting in some wasted gas.