Purpose

Ultrashort T₂^* relaxation times, found e.g. in lung tissue¹ require MR imaging approaches with short echo times (TE). Ultrashort echo time (UTE) sequences enable data acquisition with minimization of TE by acquisition of the data during ramping of the read-out gradient as opposed to only sampling when the gradient plateau is reached.
To eliminate respiratory motion, retrospective gating techniques are frequently applied to lung imaging, overcoming the highly limited maximum scan duration for a breath-hold acquisition of a few seconds. In this contribution, we applied a single petal rosette trajectory (SPR), which more efficiently covers k-space by additional data sampling during the required rephrasing of a UTE acquisition. The performance of the suggested trajectory and variants thereof for self-gated free-breathing acquisitions is investigated in volunteers and compared to radial UTE imaging.

Methods

The suggested approach was tested in 5 healthy volunteers with no reported respiratory disorders, who provided informed written consent prior to the MR examination. Images were acquired during a single breath-hold, each for inspiration and expiration, as well as during a free breathing scan of 2-3 minutes at a 3 T whole-body clinical imaging system (Ingenia 3.0T CX, Philips Healthcare, Best, The Netherlands) using a 32 (16 anterior, 16 posterior) torso coil (dStream Torso, Philips Healthcare). All data were acquired in coronal slice orientation centred at the bifurcation of the trachea.
The SPR trajectory² is parametrized by
$$k(t) = k_{\text{max}} \sin(\omega t) \text{e}^{it} ,$$
with k being the position in k-space, k_max the outmost position of sampling k-space and $$$\omega$$$ the angular frequency, which was chosen as 5/3. The radial UTE acquisition followed the radial UTE center-out sampling pattern³. In both approaches, an angular increment following the tiny golden angle sampling scheme ($$$\varphi_7$$$ = 23.6281°)⁴ was used. The number of readouts for both approaches was determined to Nyquist sample k-space for the radial UTE sequence. All relevant scan parameters are listed in table 1.
For image reconstruction, an in-house built reconstruction framework, implemented in MATLAB (The MathWorks, Natick, Massachusetts, USA) was used. The k-space density functions for the SPR trajectory were calculated using a Voronoi tessellation⁵. The gating was performed using a k₀-based self-gating approach (SG_ksp) (SG signal derived from temporal evolution of DC-signal, subsequently bandpass filtered and combined with principal component analysis to get the respiratory navigator signal), as well as two image-based approaches (SG_img) (navigator signal derived from high-temporal resolution images by gradient analysis of the lung-liver interface (LLI), respiratory stages were determined by a histogram-based approach (SG_img,hist), as well as an absolute value-based approach (SG_img,abs)). The self-gating reconstructions are described in further detail in Metze et al.⁶ A temporal resolution of 100 ms was chosen for all self-gated reconstructions.
The parenchyma was segmented semi-automatically and SNR was calculated according to
$$\text{SNR} = \sqrt{2-\frac{\pi}{2} } \frac{\text{SI}_{\text{ROI}}}{\omega_{\text{BG}}},$$
with SI_ROI being the signal intensity of a region of interest (ROI) and $$$\omega_{\text{BG}}$$$ the standard deviation of background noise. Additionally, image sharpness was assessed by a standard metric⁷, identifying the positions of pixels corresponding to 25 and 75% of the maximum signal intensity of an intensity profile, chosen across the LLI.

Results

Data acquisition could be completed for all volunteers.
SNR improved by 20% for breath-hold acquisitions when comparing SPR and radial UTE images. For the self-gated reconstructions, the SNR in the lung parenchyma increased between 33% and 41% as can be appreciated in figure 2.
Figure 3 shows the line profile placed over the LLI for breath-hold and self-gated images of one volunteer for both radial UTE and SPR trajectory, for inspiration and expiration, respectively.
While sharpness for images reconstructed using SG_ksp only differs by less than 20%, the image based self-gating approaches increase sharpness by up to 270% for the expiratory phase, as can be seen in figure 4.

Discussion and Conclusion

The SPR trajectories cover more k-space data without increasing repetition time or total scan duration, since each petal ends in k₀, thus removing the necessity of a rephasing gradient applied prior to spoiling. The resulting oversampling of k-space yields improved SNR compared to radial UTE sampling.
The higher sampling density of k-space per interleave decreases aliasing artifacts in the high-temporal resolution images, thus increasing performance of especially image-based self-gating approaches, which can be appreciated in the sharpness analysis.
Considering the previously outlined compatibility of the SPR trajectory with image-based self-gating approaches, suitability of the SPR trajectory with the nuSG algorithm⁷ suggests itself to be further evaluated. Additionally, an implementation for 3D imaging within the framework of stack-of-stars might be worth pursuing.