Abstract Book

S277

ESTRO 37

OC-0524 Planning Library Based on Population Shape Analysis for Cervical Adaptive Radiotherapy B. Rigaud 1 , A. Simon 1 , M. Gobeli 2 , J. Leseur 2 , D. Williaume 2 , J. Castelli 2 , C. Lafond 2 , O. Acosta 1 , P. Haigron 1 , R. De Crevoisier 2 1 LTSI-INSERM 1099, University Rennes 1, Rennes, France 2 Centre Eugènes Marquis, Radiotherapy, Rennes, France Purpose or Objective In the context of locally advanced cervix carcinoma (LACC) ART, the aim of the study was to generate a population based modelled planning library and compare it to standard IMRT and planning library. Material and Methods The study included 19 patients receiving IMRT to a total dose of 45 Gy for LACC. Each patient had: three planning CTs corresponding to three bladder volumes (empty [EB], intermediate [IB] and full [FB]), a CT scan at 20 Gy and bi-weekly CBCTs. The CTV and OARs were manually delineated on each CT and CBCT. CBCTs were bone registered on the planning CTs. A patient specific planning library was developed to anticipate different anatomical configurations of the uterus. First, the cervix and bladder meshes of the population were registered towards a template using an in-house deformable mesh registration. Secondly, a principal component analysis of the cervix and bladder was performed to extract the dominant deformations. Then, specific deformations were extracted by constraining the movements of the tip of uterus. Using a leave one patient out procedure, the modelled planning library of three plans was built thanks to the model deformations over the IB CT. The plan of the day selection was simulated using the Dice coefficient between the CBCT CTV and the library CTVs. The modelled library was geometrically compared to a "standard" IMRT (one IB CT plan) and planning library including the three planning CTs. The criterion was the percentage of coverage of the CTV and OARs by the 10 mm PTV. Results Figure 1 represents the modelled planning library performance compared to the standard RT and the planning library in term of CTV and OAR coverage by the PTV.

Figure 1: CTV and OARs coverage by the 10 mm PTV for the strategies For the CTV coverage, no statistical difference was found between both library but both library significantly increased the coverage compared to the standard RT. Between the two libraries, the mean and min CTV coverage was increased by 1.1% and 22.5%, respectively, by the modelled approach. For the OARs coverage, the rectum and bowel was significantly sparred by the modelled planning library compared to the classic one. Both libraries significantly sparred the bowel compared to the standard RT. The bladder was significantly over-covered by the modelled library. Moreover, in case of patient showing tumor fixity before RT delivery with large per-treatment uterine motion, the modelled library anticipates more deformations (Figure 2). Thus, CTV coverage is increased where the classic planning library fail.

Figure 2: Illustration of the planning library and modelled library for a 'non-mover/mover” patient Conclusion We proposed an original approach to model a planning library based on population deformations. The modelled library appears superior to standard IMRT and planning library for increasing CTV coverage and sparing OAR. Moreover, based on only one CT, this approach allows anticipating deformations for patient with planning tumor fixity and per-treatment motion. OC-0525 An evaluation of vocal instruction for external respiratory motion using kernel density estimation C.S. Chuang 1 , R. Xu 2 , X. Li 3 , G. Royle 1 , J. McClelland 4 1 University College London, Medical Physics and Biomedical Engineering, London, United Kingdom 2 Shenzhen People’s hospital, Medical Oncology, Shenzhen, China 3 Shenzhen People’s hospital, Radiation Oncology, Shenzhen, China 4 University College London, Centre for Medical Image Computing, London, United Kingdom