ESTRO 2020 Abstract book

S975 ESTRO 2020

PO-1679 Tumor volume regression during neoadjuvant chemoradiotherapy for esophageal cancer on weekly MRI I. Defize 1 , M. Boekhoff 1 , A. Borggreve 1 , A. Van Lier 1 , N. Takahashi 2 , N. Haj Mohommad 1 , J. Ruurda 3 , R. Van Hillegersberg 3 , S. Mook 1 , G. Meijer 1 1 University Medical Center Utrecht, Radiation Oncology, Utrecht, The Netherlands ; 2 Tohoku University Graduate School of Medicine, Radiation Oncology, Sendai, Japan ; 3 University Medical Center Utrecht, Surgery, Utrecht, The Netherlands Purpose or Objective Neoadjuvant chemoradiotherapy (nCRT) for esophageal cancer causes tumor regression during treatment. Tumor regression can induce changes in the thoracic anatomy, such as smaller target volumes and displacement of organs at risk (OARs) surrounding the tumor. Adaptation of the radiotherapy treatment plan according to volumetric changes during treatment might reduce radiation dose to the OARs, while maintaining adequate target coverage. Data on the magnitude of the volumetric changes and its impact on the thoracic anatomy is scarce. The aim of this study was to assess the volumetric changes in the primary tumor during nCRT for esophageal cancer based on weekly MRI scans. Material and Methods In this prospective study, patients with a adeno- or squamous cell carcinoma of the esophagus treated with neoajduvant chemoradiotherapy according to the CROSS regimen (carboplatin and paclitaxel for 5 weeks with concurrent volumetric arc radiation therapy (41.4 Gy in 23 fractions of 1.8 Gy) were included. Of each patient, six sequential MRI scans (transversal and sagittal T2-weighted images) were acquired: one prior to nCRT, and five in each subsequent week during nCRT. Gross tumor volumes (GTV) were delineated on each MRI by two dedicated radiation oncologists. Volumetric changes were analyzed using a paired sample t-test. Results A total of 170 MRI scans from 29 individual patients were included. Tumor volume regression and changes in thoracic anatomy were clearly visible on the MRI scans (Figure 1) . The mean (± standard deviation (SD)) tumor volume at baseline was 46 cm 3 (± 23). Tumor volume regression started after the first week of nCRT and appeared to be a linear process with significant decrease in tumor volumes every subsequent week. A decrease to 41cm 3 (91% of initial volume), 37cm 3 (81%), 34cm 3 (77%), and 33cm 3 (72%) was observed in the second, third, fourth and fifth week of nCRT, respectively (Table 1) .

PO-1678 Prostate tumor characteristics in acetate-PET and MRI - Impact of androgen depravation therapy. U. Björeland 1,2 , T. Nyholm 2 , J. Jonsson 2 , L. Beckman 2,3 , K. Riklund 2 , S. Strandberg 2 , L. Blomqvist 2 , M. Skorpil 2 , C. Thellenberg-Karlsson 2 1 Sundsvalls Hospital, Radiation physics, Sundsvall, Sweden ; 2 Umeå University, Department of Radiation Sciences, Umeå, Sweden ; 3 Sundsvall Hospital, Oncology, Sundsvall, Sweden Purpose or Objective Local recurrences in prostate cancer often appear at or close to the dominant lesion within the prostate and, therefore, should be targeted with dose escalation. However, androgen depravation therapy (ADT) affects the tumor appearance in diagnostic MRI [1], and tumor lesions can be misinterpreted as normal prostate tissue. For acetate-PET, the effect on human data is lacking. In this study, we investigated the ADT impact on the ability to separate tumor and from non-tumor tissue in MRI and acetate-PET images. We used first-order statistics and texture analysis with gray-level co-occurrence matrices (GLCM) to interpret the result. Material and Methods In total, fifty patients were analyzed with MRI and acetate- PET, before and after ADT. Quantifiable images were calculated; standard uptake value (SUV) for acetate-PET, apparent diffusion coefficients (ADC) for diffusion-MRI, and volume transfer constant (Ktrans) for dynamic contrast-enhanced MRI. Tumor and a non-tumor region were identified at baseline in ADC. With rigid and deformable registrations, the regions were transferred from baseline ADC to subsequent images, see fig. 1. GLCM invariant Haralick texture features were computed along with first-order features to interpret the result. Results First-order features: At baseline, the features mean, median, standard deviation, and 95% percentile showed all significant ability to separate tumor ROI from non-tumor ROI for ADC, SUV, and Ktrans. After ADT, only ADC mean and median, and 5% percentile could separate tumor ROI from non-tumor ROI. GLCM, Haralick texture features: At baseline, several features showed a significant ability to separate tumor ROI from non-tumor ROI for ADC and SUV. After ADT, only one feature could separate tumor ROI from reference ROI: Information Measure of Correlation 2 for SUV images. See Table 1. Conclusion Using GLCM textural features to separate tumor for non- tumor in prostate cancer imaging after ADT is a challenge. The result seems to be dependent on; population, imaging parameters, and postprocessing among other variables. We were not able to reproduce the results from an earlier GLCM study on prostate ADC imaging pre-and post-ADT [2]. Pre-ADT, a few features from [2] agreed with our study. Post-ADT, we showed that ADC GLCM features could not separate tumor from non-tumor, while [2] have ten features that could separate tumor from non-tumor. Interesting to notice, post-ADT the mean value could separate tumor from non-tumor in [2]. First-order feature median or mean seems to be more reliable than GLCM to distinguish tumor from non-tumor post-ADT. For ADC and Ktrans, we could verify the median value results from [1]. In clinical practice, our suggestion is to outline the tumor ROI before ADT according to PIRADS MRI recommendations for prostate imaging [3], then transfer the ROI with deformable image registration to the images used for radiotherapy planning.