ESTRO 2022 - Abstract Book

S1377

Abstract book

ESTRO 2022

Results LEGO provided the ability to construct a highly uniform phantom to assess the geometric accuracy of CT-MR fusion across a volume similar to that used for intracranial SRS. When the phantom was correctly aligned to MR isocentre; the accuracy was <1mm centrally and <2mm at slices +/-80mm Sup-Inf. A 32mm Sup-Inf offset from MR isocentre resulted in distortions of <4mm at the edges but remained <1mm centrally. The accuracy in the central region was maintained even when the phantom was offset 64mm from the MR isocentre, but became significantly worse at the phantom edges, with distortions up to 7mm. These edge distortions would be significant, since typical dose gradients of up to 7Gy/mm can be achieved in SRS. For multiple targets, the isodose distribution is rarely spherical; in these cases, along certain planes the dose gradient may only be 2Gy/mm. The scanner’s default distortion correction filters were able to correct the extreme edge distortions to <1 mm, but with a loss of MR resolution. Conclusion This study highlights the ability to accurately assess CT-MR fusion accuracy for SRS with a LEGO phantom. For patients with multiple metastasis, additional care should be taken if there are lesions at extremes of the dataset. Good MR isocentric positioning of the patient and the use of distortion correction filters reduce these effects, and enables treatment of all lesions with good dosimetric precision. If patient alignment is not ideal there is a risk of significantly under-dosing peripheral targets due to the steep dose gradients used within SRS. S. Thulasi Seetha 1,2 , E. Garanzini 3 , A. Messina 3 , C. Tenconi 4 , C. Marenghi 2 , B. Avuzzi 5 , M. Catanzaro 6 , S. Stagni 6 , S. Villa 5 , B. Noris Chiorda 5 , F. Badenchini 2 , J. Panchakumar 7 , E. Bertocchi 2 , E. Pignoli 4 , R. Valdagni 2,5,8 , A. Casale 3 , N. Nicolai 6 , T. Rancati 2 1 GROW - School for Oncology and Developmental Biology, Maastricht University, Precision Medicine, Faculty of Health, Medicine and Life Sciences, Maastricht, The Netherlands; 2 Fondazione IRCCS Istituto Nazionale dei Tumori, Prostate Cancer Program, Milan, Italy; 3 Fondazione IRCCS Istituto Nazionale dei Tumori, Unit of Radiology, Milan, Italy; 4 Fondazione IRCCS Istituto Nazionale dei Tumori, Unit of Medical Physics, Milan, Italy; 5 Fondazione IRCCS Istituto Nazionale dei Tumori, Unit of Urology, Milan, Italy; 6 Fondazione IRCCS Istituto Nazionale dei Tumori, Unit of Radiation Oncology 1, Milan, Italy; 7 Politecnico di Milano, Dept. of Bioengineering, Milan, Italy; 8 Università degli Studi di Milano, Dept. of Oncology and Hemato-oncology, Milan, Italy Purpose or Objective To propose a novel method to evaluate the stability of radiomic features (RFs) extracted from multiparametric (mp)MRI sequences to variations in segmentation using in-silico contour generation. We applied the process to the whole prostate region. The stable RFs identified are then used to extract a robust subset of radiomic features. Materials and Methods To evaluate the stability of RFs to variations in segmentation, we simulated various under- and/or over-segmentation scenarios commonly seen in clinical practice using the principle of data augmentation in deep learning. The overall workflow is illustrated in Figure 1.a. The manual delineation by a single radiologist was perturbed using geometric transformations such as rotation, scaling, and shifting to generate the synthetic segmentations. We considered 3 categories of augmentations (in-plane, out-plane, and in&out-plane) and 2 types of biases (random and systematic) as part of the PO-1595 Automated stability study on mpMRI prostate radiomics features to variations in segmentation