ESTRO 38 Abstract book

S1123 ESTRO 38

Performing a rigid registration between CT image acquired with CBCT at the eighth fraction (last treatment's fraction) and CT Untagged, a HU Profile has been plotted referred to PTV's centroid of in both images as seen in Figure 1.

Since qualitative profile depicts analogous behaviour between both images, statistical analysis has been performance to quantify whether or not are correlated. Normality tests were performed to decide the best fit model. Attempt to correlate CBCT and Untag profiles was carried out by performing and ordinary least square (OLS) linearity model with Python 3.7. This analysis was done for each individual and also considering 'all in one' bulk data. R2 for individuals ranged from 0,736 to 0.980, yielding an average R2=0.872 for individuals (σ=0.074) and R2=0.739 for bulk analysis. In fact, 75% of patients had an R2>0.850, showing a strong correlation that only fails in HU ranges linked to air and intrinsic variability of acquisition process as seen in Quantile Quantile Plot (Figure 2).

Conclusion The accuracy of the SPR estimation in the stoichiometric method could be improved by changing the CT number estimation method. For the optimal method, the SPR accuracy (RMSE = 0.5%) was close to values obtained for dual energy CT. EP-2045 Tumor profile matching at the end of 8x7,5 Gy SBRT treatment: CBCT vs Untagged Image Reconstruction G. Pozo Rodriguez 1 , P. Garcia 2 , A. Ferrando 1 , M. Leonor 1 , A. Gaitan 1 1 Hospital Universitario 12 de Octubre, Medical Physics Department, MADRID, Spain ; 2 Hospital Universitario de La Princesa, Medical Physics Department, Madrid, Spain Purpose or Objective To assess and quantify correlation between tumor images by comparing Hounsfield Unit (HU) profiles of 4DCT acquisition and CBCT image acquired at last therapy fraction. Material and Methods A cohort of 15 NSCLC patients treated with SBRT and dose administration delivered in 8 fractions of 7,5Gy (BED=105Gy) was selected for this study. SBRT treatments were performed on a Varian Clinac 2300 iX (Varian Medical Systems, Palo Alto.CA). This linear accelerator includes an On Board Imager (OBI) unit to perform planar or CBCT KV image, with a aSi digital panel detector model PaxScan 4030CB. In each fraction, 2 CBCT are acquired for patient positioning and intra-fraction movement tracking, with an average acquisition time ( t acq) of 13,02s (σ=0.21s). 4DCT was acquired with a Philips Brilliance Big Bore CT of 64 detector rows (Philips Healthcare, Amsterdam.The Netherlands), using a nominal slice thickness of 2mm. Respiratory cycle was binned into 10 phases and ITV was generated by means of merging 10 contoured GTV. This structure is expanded 5mm isotropically and determines PTV to be treated. CT utilized for planning is an averaged reconstruction called Untagged Series. Image analysis has been performed with Image Registration tool implemented in ARIA R&V System (Varian Medical Systems) Results

Conclusion CBCT and planning Image matching in lung has been evaluated in terms of HU profile since acceptable contrast can be obtained. Focusing on target shape, in despite of apparent artifacts frequently seen in CBCT, profile analysis exhibits good correlation between initial and final image (R2=0.872). Apparently, HU topography of the tumor represented is not device-dependent as averaged target is displayed in both cases: Untagged CT and CBCT acquisition (t aq =13,02s evolving several respiratory cicles), although inherent noise may be present (OAR proximity, respiration). Even when considering bulk data analysis, an R2 of almost 0,750 reveals a predictible model at least concerning to tumor surrounding. EP-2046 Patient setup verification using synthetic DRRs in an MR only workflow for head and neck cancer E. Palmér 1 , A. Karlsson 1,2 , F. Nordström 1,2 , K. Petruson 3 , M. Ljungberg 1,2 , M. Sohlin 1,2 1 University of Gothenburg, Department of Radiation Physics- Institute of Clinical Sciences- Sahlgrenska Academy, Gothenburg, Sweden ; 2 Sahlgrenska University Hospital, Department of Medical Physics and Biomedical