Stefano Mandija1,2, Federico D'Agata1,2,3, Robin J.M. Navest1,2, Alessandro Sbrizzi1,2, Cornelis P.J. Raaijmakers1,2, Rob H.N. Tijssen1,2, Marielle E.P. Philippens1,2, Enrica Seravalli1, Joost J.C. Verhoeff1, Jan J.W. Lagendijk1,2, and Cornelis A.T. van den Berg1,2
1University Medical Center Utrecht, Utrecht, Netherlands, 2Utrecht University, Utrecht, Netherlands, 3University of Turin, Turin, Italy
Synopsis
For
radiotherapy treatment planning, it is essential to perform the MRI/CT exams in
treatment position. For this purpose, thermoplastic immobilization masks are
used for brain and head-and-neck radiotherapy. However, since standard
immobilization masks do not fit in the diagnostic MR head/neck coils,
suboptimal surface-loop coils leading to poor image quality are used in
clinical practice. Here, we present a new immobilization setup. This setup has
several advantages compared to state-of-the-art setups: it fits in the diagnostic
head/neck MR coils, it allows diagnostic image quality in treatment position,
high SNR, homogenous signal, restricted motion (about 1 mm) and accurate
inter-fraction repositioning.
Introduction
In
Radiotherapy (RT) clinical practice, the superior soft tissue contrast of MRI
images compared to CT makes MRI an essential image modality. MRI provides
multi-contrast visualization of tumors, which allows tumor delineation and
monitoring of RT treatments.
For
RT treatment planning, it is essential to perform the MRI and CT
exams in treatment position. For this purpose and to minimize
inter/intra-fraction movement, thermoplastic immobilization masks are used for
brain and head-and-neck (HN) RT. However, standard immobilization masks do not
fit in the diagnostic MR head/neck coils. As a compromise, flexible surface
coils are adopted1-6 despite their technician dependent positioning and
inferior signal-to-noise-ratio (SNR)7 compared to diagnostic head/neck
coils. This leads to relatively poor image quality and reproducibility. The
goal of this work is to create a new immobilization setup that fits into the
diagnostic MR head/neck coils thereby boosting MR image quality and
reproducibility compared to standard RT setups.Materials and Methods
For
the standard and the novel RT brain/HN setups (Fig. 1), MR images were acquired
on 2 volunteers using a 3T Ingenia MRI (Philips Healthcare, Best, The
Netherlands).
The standard RT setup consisted of an individualized
five-points head-and-shoulder mask fixated to the MRI flat table top, a
standard neck support, and two MR flexible surface receive coils (diameters = 15
cm for HN, 20 cm for brain).
The novel RT setup consisted of an individualized three-points
mask (for brain) or three-points head-and-shoulder mask (for neck) that fit
into the diagnostic MR head/neck coils. A new fixation thermoplastic base hosting
the anchor points for the immobilization masks was created to fit in the MR head/neck
base. The proposed setup can be used for both 1.5T and 3T scanners given the
same base geometry.
Several
comparative tests were performed (Table 1).
-
SNR test: relative SNR maps were computed
from 2D T1-weighted Fast-Field-Echo scans (2 dynamics, the second being a noise
scan)8.
- Image quality test: 3D T1-weighted Turbo-Field-Echo,
3D T2-weighted Turbo-Spin-Echo FLAIR and 2D T2-weighted Turbo-Spin-Echo scans
were acquired for brain, for neck 2D T1-weighted Turbo-Spin-Echo and 2D T2-weighted
Turbo-Spin-Echo mDIXON scans were acquired.
-
Motion restriction test: the maximum motion in the
feet-head/left-right directions was estimated from 2D cine-MR T1-weighted balanced-Fast-Field-Echo
acquisitions (300 dynamics) using an Optical Flow algorithm9.
-
Inter-fraction repositioning test: two high resolution 3D T1-weighted
Turbo-Field-Echo sequences were acquired for each setup. Between acquisitions,
the immobilization mask was removed and the subject was asked to move. The mean
and standard deviation of the subject movement between each pair of MR acquisitions
were computed using Optical Flow as a proxy of the reproducibility of
inter-fraction subject repositioning.
Results
The
proposed setup allowed between 2 and 3 times higher SNR values for brain MRI
(Fig. 2-left). For the standard RT-HN setup, the SNR was highly dependent on
the spatial position (Fig. 2-right). Only locally, below the flex coils, the
standard RT-HN setup had comparable SNR to the proposed RT-HN setup (see
transverse images). This was not the case for deeper located anatomical regions,
e.g. throat, chest and shoulders (sagittal images), where the SNR of the
proposed RT-HN setup was about 3 times higher.
Brain
images acquired with the standard RT setup had an inferior diagnostic quality
compared to the new setup, especially for the FLAIR (inferior sensitivity in
detecting small lesions, Fig. 3).
For
HN, the image quality obtained with the standard and the proposed setups was
comparable in the regions below the flex coils. However, it can be expected
that the image quality is reduced for the standard HN setup in regions away
from the flex coils (e.g. low neck and shoulders) due to the reduced SNR.
Comparable
motion restriction in feet-head/left-right directions (maximum motion≈1mm, Fig.
4) and comparable inter-fraction repositioning accuracy (mean inter-fraction movement
1±0.5mm) were observed for the standard and the novel setup.
Discussion and Conclusions
We
created an immobilization setup that fits in the diagnostic head/neck MR coils.
This setup has several advantages compared to state-of-the-art setups:
diagnostic image quality in RT treatment position, high SNR, homogenous signal,
restricted motion (about 1 mm) and accurate inter-fraction repositioning. The
presence of more receive elements in the head/neck MR coils compared to the
flex coils offers more opportunity for parallel imaging. Furthermore, lack of
robustness related to the use of surface flexible coils is avoided, e.g.
coupling due to technician dependent coil positioning, which often results in relatively
poor image quality. The translation of the new setup to the treatment table
will be further investigated.Acknowledgements
The authors would like to thank Fred Groen, Tuan Nguyen, Ellart Aalbers and Teun Coolen for their assistance and support.References
1) Hess CF, et al Radiother Oncol (1995), 34:69–72.
2) Hanvey S, et al. Phys Med Biol (2009), 54:5381–94.
3) Verduijn GM, et
al. Int J Radiat Oncol Biol Phys (2009), 74:630–6.
4) Ruytenberg T,
et al. Front Oncol (2018), 8:216.
5) Wong OL, et al.
Quant Imaging Med Surg (2017), 7:205–214.
6) Winter RM, et al. Radiother Oncol (2018),
128(3):485-491.
7) Hayes CE, et al. Med Phys (1985), 12:604–7.
8) Kellman
P., et al. MRM
(2005), 54:1439-47.
9) Zachiu C., et
al. Phys Med Biol (2015), 60:9003–29.