Three dimensional inversion recovery dual-echo ultrashort echo time imaging with k-space reordering for effective suppression of longer T2 species in lung parenchyma imaging
Neville D Gai1, Ashkan A Malayeri1, and David A Bluemke1

1Radiology & Imaging Sciences, NIH, Bethesda, MD, United States


Effective imaging of short T2 species requires efficient suppression of longer T2 tisues to maximize short T2 contrast and dynamic range. While inversion with segmented k-space acquisition in Cartesian schemes is straightforward, inversion with segmented k-space UTE radial acquisition offers some challenges since the center of k-space is sampled with each acquisition resulting in magnetization modulation related artifacts. Here we perform 3D inversion recovery dual-echo UTE imaging of lung parenchyma using a reordered k-space radial scheme to perform artifact free high contrast imaging of native lung parenchyma.


To develop a technique for obtaining high contrast images of native lung parenchyma using ultrashort echo time imaging.


Effective contrast agent free imaging of short T2 species requires efficient suppression of longer T2 species to maximize short T2 contrast and dynamic range. Previously, 2D dual inversion recovery UTE imaging[1], dual-echo inversion recovery UTE[2,3] and long T2 saturation UTE[4] have been employed for high contrast cartilage, bone, tendon and myelin imaging. Most of the techniques are 2D and acquire one radial line per inversion or saturation pulse and work well in areas with limited motion such as head or knee but are not suitable for lung imaging. In this work, we design a scheme for effective suppression of longer T2 species in the lungs using an inversion recovery dual echo ultrashort echo time (IR-DUTE) 3D frame work with respiratory triggered (RT) segmented acquisition.


A hyperbolic-secant adiabatic inversion pulse with the following parameters was utilized to saturate short T2* lung while inverting spins from the longer T2 species: B1max = 13.5µT, µ=4.35, b=969 rad/s, pulse width = 18ms, BW = 1.34kHz. The offset frequency was set to 220Hz to cover both fat and water peaks[2]. Bloch simulations were performed to ascertain magnetization from lung, fat, muscle and blood. Since inversion pulse was targeted for fat, TE2 need not be chosen to be in-phase with TE1 and was fixed based on conflicting requirements of higher SNR and lower blurring[5]. Subtraction of the second echo image from the first provides a theoretical (Bloch simulation) reduction in fat, muscle and vessel signal of 98%, 97% and 99% while reducing lung signal by just 14%. The spokes in different segments were reordered in a Modulo 2 forward-reverse scheme to provide near symmetric modulation of magnetization and were compared with a standard sequential acquisition scheme. The modified 3D sequence was used to scan six volunteers on a 3T Philips Achieva scanner (version 3.2.3 software). The following imaging parameters were used: FOV=38 cm, TR/TE1/TE2=4.2/0.1/1.7 ms, θ = 5.5°, TI=320 ms, 128 spokes/segment, 512 spokes, RT, res = 1.5x1.5x5mm3, ~35 slices, scan time: ~8:30. The contrast between lung and fat and lung and muscle was assessed by drawing ROIs in the relevant tissue type in IR-DUTE and corresponding UTE images. Positive lung CNR was measured as (SNRL – SNRF,M).


Normalized Mz values from Bloch simulations were 0.24 (lung), 0.11 (fat), 0.27 (muscle) and 0.21 (blood). The point spread function (central 50 mm FOV) for fat before and after reordering of spokes from sequential to a Modulo-2 forward-reverse scheme is shown in Figure 1. Comparison first echo phantom and in-vivo images after inversion for the two ordering schemes are shown in Figures 2 and 3. The linear ordering scheme results in noticeable artifacts from fat compared to the acquisition scheme devised to provide uniform modulation of magnetization. Figure 4 shows two example slices after subtraction of echoes and provides enhanced lung parenchyma signal as expected. Mean (std) of CNR values before and after contrast manipulation are given in the Table 1.


Muscle and fat suppression can be helpful in better characterization of lesions that are close to the chest wall or mediastinum. In addition, fat suppression in the chest wall decreases artifacts related to patient motion. Effective suppression of a tissue in segmented acquisition is easily achieved in Cartesian imaging but may result in artifacts in segmented radial imaging as shown here. This is because the center of k-space is well defined in Cartesian acquisition by one phase encoding line whereas each k-space acquisition begins at the center for UTE radial imaging. A golden angle scheme would not be effective with the limited number of shots per kz encoding used in RT acquisition here. Lung parenchyma has roughly 1/10th the proton density of muscle [6]. As a result, subtraction targeted to muscle but inversion targeted to fat still yields a low lung-muscle CNR. Fat suppression could be employed but does not achieve the same efficacy when the segment duration is long as is the case here. Relative motion between the two echoes can result in slight misregistration between the subtracted images so that finer structures such as vessels may not be effectively suppressed.


A 3D segmented inversion recovery dual echo time technique with k-space reordering was shown to achieve effective long T2 and fat suppression for native lung parenchyma imaging. Full coverage of lungs with respiratory triggering can be achieved in a clinically feasible time.


No acknowledgement found.


[1] J. Du et al. MRM 2010;63:447-455 [2] C. Li et al. MRM 2012;68:680-689. [2] J. Du et al. Neuroimage 2014;87:32-41. [4] P. Larson et al. MRM 2006;56:94-103. [5] J. Rahmer et al. MRM 2006; 55:1075-1082. [6] R. Mulkern et al. Con. Mag. Res. A 2014; 43:29-53.


Table 1: CNR values for lung-fat and lung-muscle with 3D UTE imaging and with the reordered k-space IR-DUTE technique.

Figure 1: Profile of the PSF through the center of FOV (central 50 mm) simulated for fat with a standard linear ordering scheme and with the described reordered k-space scheme.

Figure 2: Image obtained with a standard linear scheme (left) shows artifacts when compared with the reordered radial k-space scheme (right).

Figure 3: First echo in-vivo images obtained 3D IR-DUTE and the two k-space ordering schemes. The one on left corresponds to linear k-space scheme while the one on right is a reordered Modulo-2 forward-reverse radial k-space sampling scheme.

Figure 4: High contrast lung parenchyma images (two slices) obtained after inversion of magnetization and subtraction of the two echo images.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)