ESTRO 37 Abstract book

S1088

ESTRO 37

planar imaging sequence with b -values of 0- 1000 s/mm 2 , and TR/TE = 3000/120 ms provided data for Apparent Diffusion Coefficient (ADC) maps. 41 lesions were contoured and labeled according to respective patient gleason score. Radiomics features extracted from 2 manual contours and 3 thresholds contouring algorithm (range 120-140% of the minimum value) were correlated with gleason score. A predictive model was developed by selecting features able to discriminate gleason 6 vs higer grade lesions among all the 5 contouring styles(fig1).

EP-1996 A patient-specific tumor control probability model based on total lesion glycolysis of anal cancer V. Skingen 1 , E. Rusten 1 , B. Rekstad 1 , C. Undseth 2 , M. Guren 2 , E. Malinen 3 1 Oslo University Hospital, Department of Medical Physics, Oslo, Norway 2 Oslo University Hospital, Department of Oncology, Oslo, Norway 3 University of Oslo, Department of Physics, Oslo, Norway Purpose or Objective Anal cancers (AC) show pronounced uptake of 18F- fluorodexoyglucose (FDG) in positron emission tomography (PET). The purpose was to develop a patient- specific tumor control probability (TCP) model employing the PET-based total lesion glycolysis (TLG), to assess its prognostic role and to use the model to estimate the gain in local control from dose escalation (DE). Material and Methods Eighty-eight patients with AC receiving conventional radiotherapy (RT) were prospectively included. Patients were regularly followed up at 3-6 months intervals after therapy, and only local recurrences were scored as events. FDG-PET was done prior to therapy. A TCP model was developed based on 1) published intrinsic radiosensitivity and fractionation sensitivity for AC and 2) TLG reflecting the number of clonogenic cells within the gross tumor volume (GTV). Furthermore, a scaling factor was used for the model to provide a global population- based control level of 80% based on previous literature. The model was thus developed without use of outcome data for the given cohort. The impact of DE from 58 Gy (conventional treatment) to 65 Gy (proposed new treatment) was assessed by the TCP model. Results Fourteen patients had local recurrence. The median and range of the patient-specific TCPs was 0.73 (0.08, 1.00). Dividing patients into two groups with low and high TCPs, the local control levels were 56 % and 91 %, respectively (Figure 1). This difference in recurrence-free survival was highly significant (P=0.001; Cox regression). A comparable TCP model based on GTV did not show any association with recurrence rates. The median increase in estimated TCP from DE was 0.52 and 0.23 for the group with low and high TCPs from conventional RT, respectively. Conclusion The patient-specific TCP model incorporating TLG was predictive of local recurrence, and was superior to a comparable model incorporating GTV only. The model estimated a clear benefit of DE for patients with high- pretreatment TLG and thus low TCP from conventional RT dosage. EP-1997 Monte Carlo simulations of direct DNA damage on gold nanoparticle enhanced proton therapy M. Sotiropoulos 1 , N.T. Henthorn 1 , J.W. Warmenhoven 1 , R.I. Mackay 2 , K.J. Kirkby 1,3 , M.J. Merchant 1,3 1 University Of Manchester, Division of Cancer Sciences, Manchester, United Kingdom 2 The Christie NHS Foundation Trust, Christie Medical Physics and Engineering, Manchester, United Kingdom 3 The Christie NHS Foundation Trust, Manchester, United Kingdom Purpose or Objective Gold nanoparticles have demonstrated a radiosensiti- zation potential under proton irradiation. Initially the radiosensitization effect was attributed to physical interactions of radiation with the gold and the production of secondary electrons that induce DNA damage. However, experiments have revealed that biological and chemical mechanisms may have significant contribution to the DNA damage and the radiosensitization effect.

Fig1 Different contouring methods, 2 manual contouring, 120%, 130%, 140% threshold contouring of lesion and healthy tissue Results Correlation with gleason 6 versus higher grade was found for one feature for the manual contouring(fig2). Four more feature agreed with the main endpoint employing the automatic contouring in the threshold range. A model based on the feature selected by manual contouring scored an AUC of 0.79 whereas the model based on threshold algorithm an AUC of 0.85.

Fig2 Grey level run length matrix - Short Run Emphasis feature boxplot for 2 manual contours in blue and green. Dashed red line shows the threshold of the model. Conclusion Our study underlined the critical importance in the choice of contouring style. Agreement of 4 feature correlation with gleason score was confirmed for semi automatic algorithm in a wide threshold range scoring a high predictive power. Validation with 8 external patient is ongoing.