Geometric distortions in clinical MRI sequences for radiotherapy: insights gained from a multicenter investigation

Figures & data

Table 1. MRI/MRL scanners and hardware specifications.

Center	Name	MRI scanner	B0 [T]	Bore size [cm]	Gradient amplitude [mT/m]	Gradient slew rate [T/m/s]
1	MRI1	Philips Ingenia	3.0	70	45	220
2	MRI2	Philips Ingenia	3.0	70	45	220
3	MRI3	Philips Ingenia	1.5	70	45	200
4	MRI4	Philips Ingenia	1.5	70	45	200
	MRL1	Elekta Unity	1.5	70	15	65
5	MRI5	Siemens MAGNETOM Aera	1.5	70	45	200
6	MRI6	Siemens MAGNETOM Vida	3.0	70	60	200
	MRL2	ViewRay MRIdian	0.35	70	18	200
7	MRI7	Siemens MAGNETOM Sola	1.5	70	45	200
	MRL3	ViewRay MRIdian	0.35	70	18	200
8	MRI8	Philips Ingenia	1.5	70	45	200
9	MRI Rad	Philips Ingenia	3.0 T	70	45	220

Figure 1. Baseline geometric distortions. The 95th percentile maximum distortion within a spherical volume with a diameter of 10 cm (DSV10, blue), 20 cm (DSV20, red), 30 cm (DSV30, yellow), and 40 cm (DSV40, purple) around MRI isocenter are shown. Scanners are grouped in field strength. σ denotes the standard deviation across all scanners.

Figure 2. Magnitude distortion (in mm) within a spherical volume with top) a diameter of 10 cm (DSV10) and bottom) with a maximum diameter Determined by the field-of-view (FOV) of the MRI sequence (DSVmax) for all clinical MRI sequences investigated. DSVmax for (brain, head/neck (H/N), abdomen, pelvis) = [30, 30, 40, 35] cm. The MRI sequences are divided into anatomical sites (brain, H/N, abdomen, pelvis) and sequence weightings representative for that anatomical site (e.g., T1w) on the x-axis. Marker color specifies MRI system and marker type specifies MRI sequence types (gradient-recalled echo (GRE), spin-echo (SE), and echo-planar imaging (EPI)). σ denotes standard deviation within a sequence group.

Figure 3. A Stereotactic dose plan for graphical illustration of the effect of image created in the treatment planning system pinnacle (Philips Healthcare, The Netherlands) on a CT reference dataset (top), where an off-axis sphere was delineated as a gross tumor volume (GTV, red) in the axial, sagittal and coronal plane. A planning target volume (PTV, cyan) was created by a 2 mm isotropic margin expansion of the GTV. The goal of the single arc volumetric modulated arc therapy (VMAT) dose plan was to cover the PTV by 30 Gy conformally while keeping steep dose gradients outside the target. One of the acquired pelvic MRI sequences (2D T2w SE, ID 92) (bottom) with a high distortion level was registered to the CT based on the central spheres. Iso-dose curves of 30 Gy (dark blue) and 20 Gy (magenta) are shown on CT images (top). the coverage of the PTV was compromised due to the distortions. Image distortion in the cranio-caudal direction are evident in this image data set, as also seen by the outer shape of the phantom.

Figure 4. Influence of MRI sequence parameters on geometric distortion. (a) Magnitude distortion as a function of pixel bandwidth (pBW/B0) for 2D sequences. (b) Magnitude distortion as a function of echo time (TE). (c,d) same plots for 3D sequences. (e,f) Same plots for diffusion-weighted (DWI) sequences. Blue: distortion within a spherical volume with a diameter of 10 cm (DSV10). orange: distortion within DSV20. Yellow: distortion within DSV30. TSE: turbo spin-echo, GRE-EPI: echo planar imaging gradient-recalled echo (Siemens resolve sequence), SE-EPI: echo planar imaging spin echo. Dashed lines show trend lines.

Figure 5. Boxplot showing distortion levels for 2D and 3D T1w/T2w/true FISP and 2D diffusion-weighted (DWI) MRI sequences within a spherical volume with a diameter of 30 cm (DSV30). the red line shows the median value. n denotes number of MRI sequences within each group.

Data availability statement

The raw image data of this study are available from the corresponding author, SWH, upon request. The authors confirm that the data supporting the findings of this study are available within the article’s supplementary materials (Table S2 and S3).