Pulmonary imaging is of clinical significance. Currently pulmonary imaging is mainly performed with CT, which raises concern of ionization radiation especially in the young. Pulmonary imaging with MRI is desirable but challenging due to the lower proton density, short T2* in lung parenchyma, and the respiratory motion. Recently, Ultra-short TE (UTE) MRI has been shown to perform similarly as standard and low dose CT in diagnosing pulmonary diseases owing to its short TE capability and resilience to motion (1). Lung parenchyma T2* values measured with a single exponential model have also been demonstrated to have a role in pulmonary functional loss assessment (2). However, the single compartment model does not fully describe the signal evolution especially at longer TEs. In this work, we propose to improve the characterization of T2*s in the lung parenchyma with a bi-exponential model. Using a 3D multi-echo radial sequence, our results demonstrated that short T2* (T2*S) values and the volume fractions of the two compartments could be readily obtained on a clinical 3T scanner.

Pulmonary imaging was performed on two volunteers on a 3T Toshiba Vantage Titan 3T scanner (Otawara, Japan) with a four-half-echo 3D radial sequence (Figure 1). Images were acquired during the expiration phase with respiratory-gating. To minimize the impact of eddy currents, a fly-back readout was used. The echo space was set to 1.1ms so that the short term eddy currents (time constant tc ~100us) induced by the gradients before the next readout decayed to a negligible level (~5tc). Three acquisitions were carried out on each volunteer with the TEs of the first echoes set to 0.1ms, 0.5ms and 0.9ms, respectively. Other acquisition parameters include TR/flip angle/FOV/slab thickness/resolution/# of radial lines/# of segments: 5.8 ms/6°/FOV 50cm/25.6cm/3.9 mm isotropic/18432/64.

Image reconstruction was performed using gridding followed by complex Fourier transform. Eddy currents during the readout was accounted for with the measured actual trajectories (3). All 12 echo images were used for nonlinear bi-exponential fitting in Matlab (Mathworks Inc.) on a voxel by voxel basis. T2*S range was set to be 0.1ms-10ms, and long T2* (T2*L) was set to be 5ms -1000ms for the fitting.

Typical signal evolutions in several typical voxels are shown in Figure 2. Voxel #1 was selected in liver tissue and the signal was nearly flat (not shown). At least two compartments are visible in plots of most lung parenchyma voxels (#2-#5), especially voxels #4 and #5. The bi-exponential model fits the data fairly well, although a single exponential model may also work for other tissues such as the vessels and the lung parenchyma voxels with high water content (e.g., #6). A small peak is observed in some plots, which may be due to the chemical shift of pulmonary surfactant.

The calculated T2*S maps and the volume fraction maps of the short and long T2* components are shown in Figure 3. Lung parenchyma tissue show higher T2*S values and volume fractions than other tissues. The calculated T2*S values of lung parenchyma range from 0.4-1.1 ms, consistent with those reported in literature. The fractions of T2*S values are different at the anterior and posterior sides, which increased from ~50% to ~70% in the ROIs placed in lung parenchyma from anterior to posterior. T2*S fraction values were low (below 20%) in other tissue such as liver, kidney, as expected.

The origin of multiple compartments in lung parenchyma voxels could be due to the existence of multiple tissue types in lung parenchyma. However, the partial volume effect of lung parenchyma with blood could also contribute to the observed phenomenon. In the latter case, the bi-exponential model may help to differentiate the contributions from lung parenchyma and blood and consequently increase the accuracy of T2*S measurement.

There are several limitations in our studies. The accuracy of the T2*L needs further investigation due to the relatively short TEs used. As can be seen in the plots, chemical shift could also compromised the accuracy of T2* characterization. Inclusion of chemical shift in modelling could help. In addition, all examinations were performed in the supine position in this study. Positional difference needs to be further investigated.

We have demonstrated the application of a multi-echo 3D radial sequence with UTE capability to image lung parenchyma on healthy volunteers. Our data indicate the existence of multiple compartments in lung parenchyma voxels. In addition to the improved accuracy of T2* measurement in lung parenchyma with the two-compartment model, the added fraction values of the two compartment could also potentially be used as biomarker for lung diseases.