ESTRO 2020 Abstract book

S1011 ESTRO 2020

area. The exact motion and dosimetric effects are thus yet uncharted territories in radiotherapy. Our study aim was to develop a technique based on high quality 3D MRI movies to visualize and quantify bowel motility and apply it in a cohort of gynecological cancer patients. Material and Methods We developed a MRI acquisition suitable for 3D motility quantification: a balanced turbo field echo sequence with TE=1.39ms and TR=2.8ms, acquiring images in 3.7 seconds (dynamic) with a 2.5mm isotropic resolution, yielding a field-of-view of 200x200x125mm 3 . During a 10 minute scan 160 dynamics were acquired. Each dynamic was deformably registered to its preceding dynamic using a B-spline transformation model, resulting in 159 3D deformation vector fields (DVFs) per MRI set (Figure 1), describing the magnitude and direction of motion during the scanning time. From the 159 DVFs the average vector length (AVL) and the length of the average vector (LAV) were calculated per voxel. Both measures were used to generate motility maps. From the maps the most suitable measure was determined to quantify motility and to be introduced as cumulative motility volume histogram (MVH) of the bowel bag volume. Finally, interpatient variation of bowel motility was analyzed using MVH parameters. Results For this study, fifteen motility MRIs of gynecological patients were analyzed. The AVL motility mapping results in a clear visualization of areas with small and large movements in contrast to the LAV measure (Figure 1). The results of the LAV based motility maps show a minimum bowel movement, indicating that a LAV value close to 0 refers to peristaltic movement (Figure 1). Therefore the AVL is used to generate the MVH for further analysis of bowel motility. Figure 2 shows the interpatient differences in bowel motility, based on the cumulative MVHs of the bowelbag volume (excluding the uterus) and the calculated MVH parameters M2cc, M10%, M50% and M90% for all patients.The M2cc is 8.2mm (range 4.7-13.6mm) and M10% is 5.2mm (range 2.7-8.2mm) on average over all patients. The mean M50%, representing the median bowel motility per 3.7 seconds, is on average 2.5mm (range 1.2-5.3 mm) and the mean M90% is 0.6mm (range 0.4-1.3mm). Conclusion We have developed a method to visualize and quantify bowel motility. In addition, we introduced the concept of MVHs, which seems to be a promising tool to quantify bowel motility. This cine MRI based quantification tool can be used in further studies to determine the effect of EBRT and BT on bowel motility and consequently the dose to the bowel.

PO-1731 Deep learning based conversion of CBCT to synthetic CT for prostate radiotherapy P. Zaffino 1 , R. Raso 2 , M.C. Angiocchi 2 , M. Merola 1 , S. Canino 2 , M. Nonnis 2 , A. Bavasso 3 , C. Mezzotero 3 , R.A. Anoja 2 , E. Mazzei 3 , M.F. Spadea 1 1 Magna Graecia University, Experimental and Clinical Medicine, Catanzaro, Italy ; 2 Azienda Ospedaliera Pugliese-Ciaccio, SOC Fisica Sanitaria, Catanzaro, Italy ; 3 Azienda Ospedaliera Pugliese-Ciaccio, UOC Radioterapia oncologica e radiobiologia, Catanzaro, Italy Purpose or Objective The aim of this work was to convert CBCT images into synthetic CT (sCT) by a deep learning algorithm, in the vision of online adaptive radiotherapy. Material and Methods 20 prostate cancer patients were retrospectively enrolled in this study. For each patient, the initial planning CT (pCT) and the first daily CBCT were matched by deformable image guided registration, in order to compensate for internal organ movements. CBCTs were converted into sCTs by a deep learning algorithm previously validated for MRI to CT conversion. Briefly, 3 convolutional neural networks were independently trained to predict the Hounsfield units in the axial, sagittal and