ESTRO 2022 - Abstract Book

S1416

Abstract book

ESTRO 2022

Conclusion This method provides a first step towards allocating MI to material specific dose kernel by estimating their Z eff from dual energy topograms. Increased mAs impact on imaging plate gain factors for uncertainty reduction and adding MI in anthropomorphic phantoms should be further investigated.

PO-1625 Modeled clinical target volume in soft tissue using diffusion weighted MRI data

N. Shusharina 1 , T. Bortfeld 1 , J. Coll-Font 2 , C. Nguyen 3

1 Massachusetts General Hospital, Radiation Oncology, Boston, USA; 2 Massachusetts General Hospital, Cardiology, Boston, USA; 3 Massachusetts General Hospital , Cardiology, Boston, USA Purpose or Objective In soft tissue sarcoma (STS), microscopic tumor spread follows the pattern of local invasion into tissues, preferably along muscle fibers. Therefore, anisotropic properties of the tissues must be taken into account in order to improve accuracy of the clinical target volume (CTV) definition. Currently, radiotherapy treatment planning for STS lacks imaging modalities and algorithms to account for the tissue anisotropy. Here we propose to include diffusion-weighted MR sequence into pre- treatment evaluation of STS to generate data for automated delineation of the CTV. Materials and Methods Eight healthy volunteers, five men and three women participated in this study. The volunteers were scanned supine, feet first using 3T MRI system (Siemens, Magnetom Prisma, Siemens Healthcare, Erlangen, Germany) and a flat body coil covering left and right thighs. The imaging protocol consisted of (a) two high resolution anatomical MRI scans (spin-echo, SE), T1- and T2-weighted; (b) a diffusion weighted (DW) SE scan using an echo planar (EP) acquisition with fat suppression. Anatomical and diffusion weighted scans were acquired in the axial plane. The DWI acquisition consisted of two b0 image with b0=50 s/mm 2 and 12 DW images with b=400 s/mm 2 using 12 gradient directions. A fat saturation was used to suppress the fat signal (SPAIR). T1- and T2-weighted acquisitions were used to match the anatomical location of the muscles in DW images. The CTV boundary was determined by iso-distant surfaces on the map of shortest path lengths from a given voxel to the gross tumor volume (GTV). The calculation is based on graph-type search on the voxel grid of the image. The non-uniform CTV boundary was found by applying the fast marching method through solving anisotropic Eikonal equation. We use DW- MRI data as an input to our model. Results The tissues (femur, bone marrow, fat) and individual muscles were manually segmented on T1-weighted MR images. The DW images were pre-processed by re-slicing them to isotropic voxel size. The denoising algorithm was applied to increase the signal-to-noise ratio (SNR) of the data. The voxel-wise diffusion tensor components were reconstructed using DIPY, the imaging library in Python. Since the DWI data was acquired from healthy volunteers the GTV was modeled by a sphere randomly placed in the image volume. For the eight subjects, we have calculated model CTV by anisotropic expansion of the GTV. By comparing the modeled CTV with established clinical guideline, the optimal scaling parameters for diffusion tensor components were determined. Conclusion Our feasibility study with healthy volunteers shows the promise of diffusion weighted MRI data for automated generation of model anisotropic CTV in soft tissue. We plan to acquire imaging data from soft-tissue sarcoma patients and apply our method to automatically define the CTV and validate the results by comparing with manually defined CTV.