ESTRO 2023 - Abstract Book

S163

Saturday 13 May

ESTRO 2023

Diffusion weighted (DW) MRI is a promising tool to assess early tumor response in radiotherapy. Apparent diffusion coefficient (ADC) maps conventionally rely on echo planar imaging (EPI). These sequences are sensitive to static magnetic field inhomogeneities especially in the phase encoding direction and are usually implemented with lower resolution. This makes their use in a low field MR-Linac context challenging. We designed a method to quantify the 3D-spatial integrity of such sequences and report potential improvements using the reversed phase-encode correction method. Materials and Methods We evaluated the vendor single shot DW-EPI sequence on a 0.35T MRIdian (ViewRay) with the following standard parameters: TR=3200ms, TE=120ms, flip angle=90°, voxel 3x3x5mm3, FOV=30x30x15cm3, three orthogonal diffusion directions and five b-values (0, 200, 300, 500, 800 s.mm-2), gantry angle (GA) at 330°. We used a 10-channel head and neck coil and additionally tested a prototype head coil (4 channels). To evaluate the geometric accuracy of low-resolution sequences we printed a 3D grid (6mm thickness, 25mm spacing, PLA) defining 5^3 markers over a 10cm3 FOV. We evaluated the 3D-printing accuracy with a CT scan (Siemens, SOMATOM, 0.78x0.78x0.6mm3). The structure was immersed in water (Figure 1b-d). The marker detection was performed using a template matching approach implemented in Matlab. The distortion is defined as the Euclidean distance to the theoretical marker position after rigid registration (Procrustes method on the 27 central markers). We investigated the influence of GA, phase encoding directions, slice orientation and receiver bandwidth on the geometric accuracy. We further evaluated the performance of reversed phase-encode correction (Anderson et al., 2003, FSL implementation). The impact of this correction on ADC values was evaluated by estimating the free water ADC (expected value=1.77x10^-3mm2.s-1 at 20°C, Krynicki et al., 1978) in the phantom using the DTIFIT module We report a mean manufacturing accuracy of 0.22±0.09mm (maximum 0.51mm) applying our marker detection method on the CT scan. The mean distortion over the different settings was 0.83±0.09mm accounting for the geometric stability of the sequence (Table 1). The distortions increase in average by 0.1mm with higher b-values (Figure 1e). The correction strategy reduced the distortion for all settings (on average by -40%) at the cost of doubling the acquisition time as it combines AP and PA images. ADC accuracy is difficult to evaluate without calibrated values, however the correction method reduced the ADCs variability and is on average closer to the expected value. of FSL. Results