ESTRO 38 Abstract book

S1042 ESTRO 38

Conclusion Although the maximal TCP difference obtained theoretically in this study is lower than the one reported in [1], where a difference of 30% was obtained, it supports the reported clinical finding. A more detailed investigation will be carried out with this and an alternative model [5] which also accounts for re- oxygenation of the tumor during treatment in order to investigate whether the observed TCP difference can be predicted by existing TCP models. 1. Alite, Fiori et al. Radiotherapy and Oncology , 121, 1, 9 – 14, 2016 2. Ruggieri R. Phys Med Biol 2004;49:4811-4823. 3. Ruggieri R, Nahum AE. Med Phys 2006;33:4044-4055. 4. Ruggieri R et al. 2012, Int J Radiat Oncol Biol Phys. 2012 Aug 1;83(5):1603-8. 5. Stavreva NA et al. Med Phys. Mar;32(3):720-5. 2005 Supported by the Bulgarian National Science Fund under contract: DN 18/4 (10.12.2017). EP-1918 Active bone marrow identification in the pelvis using texture analysis of CT features E. Gallio 1 , S. Rosati 2 , C. Fiandra 3 , F. Arcadipane 3 , A. Lesca 4 , P. Silvetti 3 , G. Balestra 2 , U. Ricardi 3 , P. Franco 3 1 A.O.U. Città della Salute e della Scienza, Medical Physics Unit, Turin, Italy ; 2 Politecnico of Turin, Department of Electronics and Telecommunications, Turin, Italy ; 3 University of Turin, Radiation Oncology- Department of Oncology, Turin, Italy ; 4 A.O.U. Città della Salute e della Scienza, Neclear Medicine Unit, Turin, Italy Purpose or Objective For anal cancer patients, hematologic toxicity (HT) may influence compliance to therapy (concurrent chemo- radiation). It is crucial to implement strategies able to identify and spare active bone marrow ( aBM ). Several methods exist to locate aBM , based on PET or MR images. The aim of this study is to evaluate the feasibility of detecting aBM on CT images with radiomics. Material and Methods Five patients were included in the study. For all patients, aBM was manually drawn on PET images (pixel size: 4x4mm 2 , spatial resolution: 6 mm) and divided in three ROIs (subregions): iliac bone marrow ( IBM ), lower pelvis ( LPBM ) and lumbosacral ( LSBM ) one. For identifying aBM on CT (pixel size: 1x1mm 2 ), three classifiers were implemented. For training of the classifiers, pelvic bone marrow ( PMB ) was segmented on CT images according to Mell L.K. et al. (Int.J.Radiat.Oncol., 2006, p. 1356-1365) procedure. The aBM was selected by PMB applying k- means algorithm (k=2), removing the pixel with the highest mean intensity (corresponding to cortical bone) and overlapping with PET aBM ROIs (after deformable registration). For all elements overlapping PET aBM , a set of 36 radiomics features was calculated: 4 first-order statistical features and 32 second-order ones. For each aBM subregion, a training set was obtained by randomly selecting ¼ of elements from five slices for each subject.

The linear model, adjusted by the number of arcs, resulted to have a R 2 of 0.9. All three input parameters were significant (P<0.05). A representation of the linear fit is reported in figure 2 Conclusion These data could be used as predictive model to assist in risk evaluation and decision for appropriate technique, such as breath hold or other suitable techniques EP-1917 Variable versus conventional inter-fraction intervals in SBRT P. Stavrev 1 , N. Stavreva 1 , A. Nahum 2 , R. Ruggieri 3 , P. Tsonev 1 , D. Pressyanov 1 1 SRD Sofia University “St. Kliment Ohridski”, Scientific Research Dpt, Sofia, Bulgaria ; 2 Clatterbridge Cancer Centre, Formerly at Physics Dpt, Bebington, United Kingdom ; 3 ‘Sacrocuore – don Calabria’ Hospital, Department of Radiation Oncology, Negrar VR, Italy Purpose or Objective To investigate theoretically the maximal possible difference in local tumor control that can be achieved with SBRT regimens with conventional and with variable This study was stimulated by [1] which reported that a higher local tumor control (by 30%) was observed clinically for a SBRT regimen employing non-consecutive, i.e. irregular rather than consecutive (conventional) inter- fraction intervals. The fractionation schemes compared in [1] are T i ,1={1,2,3,4,5} and T i ,2={1,3,8,10,15}, delivering otherwise equal total dose to the tumor. They are studied here employing a TCP model accounting for tumor hypoxia [2,3,4]; this TCP model includes re-sensitization of the tumor cells through re-oxygenation during a prolonged treatment and therefore may potentially explain the improved treatment outcome of the non-conventional SBRT schedule. The parametric space of the applied TCP model was searched in order to investigate the dependence of the obtained TCP difference on the values of the model parameters. A dose per fraction of 11 Gy was assumed, consistent with [1]. Results There exist a parameter set presented in the table for which the difference in the predicted TCPs is around 20%. The TCP dependence on the repopulation rate for both schedules is shown in the figure. inter-fraction intervals. Material and Methods