To develop a technique for relatively high resolution 3D simultaneous T1 and T2* mapping of the entire lung using a single free breathing scan in a clinically feasible time.In vivo and in vitro lung studies in humans have demonstrated significant relaxation time differences between normal and pathological tissue and malignant and non-malignant abnormalities. In-vivo mapping of lung tissue is difficult due to the low proton density, respiratory motion and susceptibility effects. Nevertheless, several techniques have been proposed for human lung T1 mapping including inversion recovery Look-Locker with 2D Cartesian sampling[1,2] or with low resolution 3D Cartesian sampling[3] and inversion recovery 2D low resolution ultrashort echo time imaging (UTE)[4]. Similarly, T2* mapping has been done using fractional echo GRE[5] and 2D UTE[6].

To reduce scan time, a saturation scheme was used. Saturation was achieved using a modified four pulse WET[7] technique optimized for lung T1/T2* by performing a search using Bloch simulations and taking relaxation into account. For the saturation scheme, T1 is estimated by T1 = ̶ TS/log[{1 - S(TS)/S0}/η], where TS is the saturation time and S(TS), S(0) refers to the signal with and without saturation; η is the saturation efficiency.

A segmented 3D ultrashort echo time (UTE) dual echo radial scheme was employed for data acquisition in an interleaved fashion such that a S(0) segment shot was followed by a S(TS) segment to reduce misregistration between the acquisitions. Respiratory triggering (RT) was employed with minimum shot duration set to 3.5s to allow near complete recovery of lung tissue signal. TS was optimized based on error propagation analysis for T1 and consideration of end expiration time. T2* was calculated using T2* = ΔTE/log[S(TE₁)/S(TE₂)], where S(TE_1,2) from the S(0) acquisition images were utilized. While TE₁ was fixed at minimum, TE₂ was optimized based on considerations of SNR, TR and error propagation analysis of T2* as a function of TE₂.

Seven volunteers were scanned under an IRB approved protocol after informed consent was obtained. Scanning was performed on a 3T Philips Achieva scanner (software release 3.2.3) with the following scan parameters: FOV=34cm, TR/TE₁/TE₂=3.3/0.13/1.3 ms, res: 2.5×2.5×6mm³, 512 radial lines, 128 lines/segment, RT, TS=700 ms, shot dur = 3.5s (minimum), 37 slices, scan time:~9:30. In addition, scanning was repeated on two volunteers after a few weeks to assess repeatability. T1 and T2* values were assessed from the maps in a semi-automated fashion by drawing an ROI around the chest cavity followed by thresholding to remove chest wall and vessels. Mean T1, T2* and standard deviation over the entire lung parenchyma and across all volunteers was calculated.

The optimized WET pulse had a total duration of 18 ms with crushers. Magnetization response from Bloch simulations over a range of off-resonance frequencies and for two different T2s (lung and muscle) is shown in Figure 1 (T1 >> 18 ms). The 95% suppression bandwidth was 1.72 kHz and 2.2 kHz for muscle and lung, respectively. B1 inhomogeneity of ±20% still resulted in a BW of 2 kHz for lung tissue with this WET pulse. Figure 2 shows sample T1 maps while Figure 3 shows T2* maps obtained in a volunteer with the above technique. T1 (mean ± std) of lung parenchyma across the seven volunteers was estimated to be 917 ± 83.6 ms while T2* was 0.9 ± 0.07 ms. Mean difference in the repeated studies on two volunteers was 1.8% for T1 and 3.8% for T2*.Lung disorders such as fibrosis, edema or emphysema may involve the entire lung making a 3D technique more pertinent. Despite interleaving the S0 and S(TS) measurements, residual relative motion between the acquisitions can affect the accuracy of the estimated T1/T2* values. However, techniques using multiple inversion times would be more prone to misregistration related inaccuracies. Previous works have estimated longer times for T1[1,2] which could be related to several factors including lack of sufficient SNR with Cartesian acquisition schemes or lower signal images at inversion times close to null of lung parenchyma, partial voluming effects from vessels (which have a substantially longer T1 values) particularly at lower resolution as well as from motion. More recent work using UTE showed lower T1 values [4,8,9] and variation of T1 with the employed echo time[7] attributed to a multicomponent model of the lung. In addition to physiological factors which affect relaxation, respiratory phase has also been shown to impact quantitative values [2].We have shown that relatively high resolution 3D entire lung T1 and T2* maps can be simultaneously acquired with a single scan in a clinically feasible time.