ESTRO 38 Abstract book

S611 ESTRO 38

“acceptable” quadrant (figure 2). This was confirmed by the RO who considered the delineations clinically acceptable. Most of the plotted points for the cochlea are in quadrant “acceptable” and some points were plotted in the “poor” quadrant, however the qualitative analysis shows that most ABS delineations were acceptable (13/15). Unacceptable results for pituitary gland, optic nerves, retina, lacrimal glands and lenses were found, which is in accordance with the DSC results. Despite the low DSC and the ROC analysis for the chiasm, the RO concluded that 11/15 ABS delineations were clinically acceptable.

Inter modality Dice similarity coefficient between myelo- CT and MRI was 0.801 ± 0.026.(Average ± SD) Inter observer Dice similarity coefficient in myelo-CT was 0.873 ± 0.030. (Average ± SD) Inter observer Dice similarity coefficient in MRI was 0.836 ± 0.092. (Average ± SD) Average volume spinal cord with MRI was1.874ml ±0.408. (Average ± SD) Average volume spinal cord with Myelo-CT was 1.682 ml ± 0.221. (Average ± SD). All contours were showed in Figure. Conclusion Inter modality variations between myelo-CT and MRI fusion, inter observer variations in myelo-CT and inter observer variations in MRI were evaluated and were larger than 0.7. The difference between myelo-CT and MRI fusion is small in contouring Spinal cord. PO-1100 Validation of Atlas Based Segmentation for OAR in the brain. A. Van Nunen 1 , A. De Graaf 1 , T. Budiharto 1 , D. Schuring 1 1 Catharina hospital, Radiotherapy, Eindhoven, The Netherlands Purpose or Objective Manual delineation of organs at risk (OAR) is time consuming and possibly variable between observers. Atlas Based Segmentation (ABS) enables automatisation of OAR delineation and can reduce workload and variability. Aim of this study was to investigate the accuracy of OAR delineation in the brain when using ABS. Material and Methods Delineations of OAR (brainstem, chiasm, cochlea, optic nerves, pituitary gland, retina, lacrimal glands, lenses and eyes) were generated using the ABS module in RayStation (Raysearch Laboratories, Stockholm, Sweden). OAR were segmented in 15 patients using four different atlases containing 10 CT scans, 25 CT scans, 10 MRI scans and 25 MRI scans. These were compared to the manual delineations. All OAR were reviewed by a radiation oncologist (RO) to determine whether they were clinically acceptable. Different parameters (volumes, specificity, sensitivity and Dice Similarity Coefficient (DSC)) were analysed and a Receiver Operating Characteristic (ROC) analysis was performed. A pairwise statistical analysis was performed to determine whether larger atlases could increase the accuracy of the automatic delineation and whether MRI atlases improve the delineation of the brainstem and chiasm. Results Highest median DSC (>0.8) was found for eyes and brainstem (figure 1). The median DSC was 0.5-0.7 for the cochlea, pituitary gland and optic nerves, and 0.4-0.5 for lenses, retina and chiasm (MRI). Lowest DSC was found for the chiasm based on the CT-atlas (figure 1). No statistically significant improvement was observed for atlases with 25 scans compared to atlases with 10 scans. The delineation of the brainstem and chiasm improved significantly by using an MRI atlas.

Conclusion Accurate delineation of the brainstem, cochlea and eyes was obtained by using ABS. Automatic delineations of optic nerves, retina, lacrimal glands, lenses and pituitary gland were not clinically acceptable. Automatic delineation of the chiasm with a MRI atlas is a good starting point, however manual corrections are needed for an accurate delineation. Increasing the number of scans in the atlas did not significantly improve the results, but caused an increase in computation time. Using an MRI atlas improved the accuracy of automatic delineations of the brainstem and chiasm. PO-1101 Feasibility of PSMA PET/CT for evaluation of radiotherapy toxicity in salivary glands N. Bruin 1 , V. Mohan 2 , J. Van de Kamer 2 , A. Al-Mamgani 2 , J. Sonke 2 , W. Vogel 1 1 Netherlands Cancer Institute, Departments of Radiation Oncology and Nuclear Medicine, Amsterdam, The Netherlands ; 2 Netherlands Cancer Institute, Department of Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective Patients treated with high dose radiotherapy for head and neck cancer may suffer from a severe dry mouth (xerostomia) due to salivary gland damage. This toxicity has been difficult to evaluate with quantitative and gland- specific measurements. Functional imaging with prostate specific membrane antigen (PSMA) PET/CT is thought to indicate viable gland cells in salivary glands, and can theoretically be used to evaluate cell loss after radiotherapy. Our purpose is to investigate the technical and logistical feasibility of repeated PSMA PET/CT to evaluate salivary gland toxicity after radiotherapy. Material and Methods Seven patients have been included in an ongoing prospective study, that will include 20 head-neck cancer patients who are treated with 35 fractions of 2 Gy. Four PSMA PET/CT scans of the head-neck area were made at baseline, during treatment, and at 1 and at 6 months post treatment. Patients were scanned in treatment position using a flat table top, base, personalized mask and knee support, using a Philips Gemini TOF PET/CT scanner. A low dose of 50 MBq PSMA was administered, and the scan was acquired after an incubation period of 45 min. The field of view was from orbit to clavicles, in 2 bed positions of 6

The ROC analysis shows that all points for the eyes and 14/15 points for the brainstem are plotted in the