Steven Jackson1,2, Hanna Hanson2, David Cobben2, Kathryn Banfill2, Ahmed Salem2, Lisa McDaid1, Marcel van Herk2, and Benjamin Rowland2
1The Christie NHS Foundation Trust, Manchester, United Kingdom, 2Division of Cancer Sciences, The University of Manchester, Manchester, United Kingdom
Synopsis
Fat saturated and unsaturated T2W images of the same anatomy are required for MRI lung radiotherapy treatment planning. A prospective 'keyhole' pulse sequence that produces both image contrasts at full resolution via acquisition of a subset of FS k-space substituted into the non FS k-space was simulated. Simulated FS images with 50% of FS k-space substituted in a cross hair pattern were found to be only slightly inferior to acquired FS images by four clinicians based on scoring for delineation suitability. The prospective pulse sequence would save MR protocol time and improve the patient experience.
Introduction
Lung cancer treatment with the Elekta Unity MR-Linac is in
preparation [1,2]. T2 weighting (T2W) is the preferred contrast for MR lung
tumour delineation on radiotherapy MR planning scans, however the similar
contrast of fat and tumour make identifying mediastinal invasion and some lymph
nodes difficult on T2W images, necessitating a second fat-suppressed (FS) image
[3]. This adds significant time to the imaging protocol, prolongs patient
discomfort during the planning MR in radiotherapy treatment position and
necessitates registration of the two datasets in post-processing. We envisage a
new technique that acquires the complete non-FS k-space and a subset of FS
k-space within one pulse sequence, then reconstructs both images at full
resolution using a keyhole k-space approach [4], offering significant time
saving during the scan, in post-processing and reduced time in the MR
scanner for the patient.Methods
A multi-slice T2W TSE pulse sequence (TR/TE = 5200/102ms, 30x3.5mm slices with matrix 320x320, 1.25x1.25mm pixels) was obtained both with
and without spectral fat saturation FS in lung cancer patients as part of
Clinical Trial ref NCT03048760 on a 1.5T Aera (Siemens, Erlangen). Re-sampling of the non-FS image data
was undertaken to ensure the imaging slices were at the same spatial locations
as the FS images. K-space data for each sequence was obtained and a keyhole
k-space matrix was created in post-processing using MATLAB (v 2016b, Mathworks,
Natick, USA). To create the keyhole, lines of FS k-space were substituted
into the complete non-FS k-space sequence in a cross hair pattern, selected because the majority of high pixel values in k-space lie close to the principal axes.
The chosen method is susceptible to ringing artefacts
generated at the boundary of the substituted k-space lines. This effect was
minimised by calculating the mean signal in the middle five lines of k-space in
the common frequency encode direction of the FS and non-FS datasets and scaling
the non-FS dataset down by the ratio of the means. The resulting keyhole
k-space was inverse Fourier transformed into a final synthesized FS image.
Synthesized FS images with 30, 40 and 50% k-space line
replacement were prepared for two slices through the visible tumour for eight
patients. 30% replacement meant 15% of k-space lines were replaced in the x-direction and the same number were replaced in the y-direction, with the overlapping centre of k-space averaged as in a radial acquisition (see figure 1 B-D, insets).
To assess the suitability of the synthesized images for
delineation 3 lung oncologists and 1 research radiographer were asked to score them
on a scale of 0-100, where the traditionally acquired FS
image was assigned a reference score of 80.
The images for scoring were presented on a 2x2 grid with the reference image in
the top left and the 30, 40 and 50% k-space line replacement images assigned
randomly to the other squares, see figure 1.Results
Figure 2 shows the results from the four scorers and averages
obtained thereafter.
The mean (standard deviation) for the images synthesized with
30, 40 and 50% k-space substitution was 66(9), 71(8) and 75(6) out of 100,
where the reference image was assigned a score of 80.Discussion
The results give confidence that the proposed keyhole FS
method can produce FS images of sufficient quality for radiotherapy treatment
planning purposes. Images synthesized via 50% FS k-space line replacement into
a full non FS k-space were found to be only slightly inferior to FS images
acquired in the traditional way.
The inter-scorer variation, as reflected in the standard
deviation of the average scores across the four scorers, was large and potentially demonstrated that oncologists look for different things from
their FS T2W planning MR images.
To prospectively acquire images using the described technique a specially designed pulse sequence would be required to capture the full non-FS image and a subset of FS k-space with two orthogonal
phase encode directions. Such a prospective pulse sequence
will be tested in further work. Beyond the simple time saving afforded by acquiring
fewer lines of FS k-space, the acquisition of an interleaved multi-contrast keyhole
pulse sequence would offer intrinsic spatial registration between the two reconstructed images, and thus add further benefit to the MR treatment
planning pathway.Conclusion
This work demonstrated the possibility for intrinsically registered fat saturated an
unsaturated T2W lung MRI images to be acquired in one keyhole pulse sequence for
the purposes of radiotherapy treatment planning. The proposed method would save time in the patient
workflow and thereby increase patient comfort and compliance. The method may
have wider clinical benefit in other anatomies for radiotherapy MR treatment
planning and in radiology where FS and non FS images are required of the same
anatomy.Acknowledgements
No acknowledgement found.References
1. Bainbridge, H. et al., 2017. Magnetic resonance imaging in
precision radiation therapy for lung cancer. Translational Lung Cancer
Research, 6(6), pp. 689-707.
2. Dubec, M. et al., 2017. MRI for Lung Cancer Radiotherapy
Treatment Planning and Treatment. Journal of Thoracic Oncology, 12(11, S2),
p. S2334.
3. Cobben, D. C. P. et al., 2016. Emerging Role of MRI
for Radiation Treatment Planning in Lung Cancer. Technology in Cancer
Research in Treatment, 15(6), pp. NP47-60.
4. Flask et al., 2003. Keyhole Dixon Method for Faster, Perceptually Equivalent Fat Suppression. Journal of Magnetic Resonance Imaging, 18 pp. 103-112.