S123
Brachytherapy - Physics
ESTRO 2026
Proffered Paper 1712
an overall time saving of 65%.
Advanced Electromagnetic Tracking with a Foley- Based Reference Sensor for Motion Compensation and Anatomical Referencing in Prostate HDR Brachytherapy Ioannis Androulakis 1 , Jeremy Godart 1,2 , Rozemarijn Ammerlaan 1 , Crystal Wang 1 , Pelagia Fotiadou 1 , Miranda E.M.C. Christianen 1 , Henrike E.M.C. Westerveld 1 , Mischa Hoogeman 1,2 , Remi Nout 1 , Inger- Karine K. Kolkman-Deurloo 1 1 Department of Radiotherapy, Erasmus MC Cancer Institute, University Medical Center, Rotterdam, Netherlands. 2 Department of Medical Physics & Informatics, HollandPTC, Delft, Netherlands Purpose/Objective: Electromagnetic tracking (EMT) enables pre-treatment error detection and implant reconstruction verification in prostate high-dose-rate brachytherapy (HDR-BT), by mapping the implant geometry [1].However patient and organ motion during EMT acquisition can distort measurements if no reference sensor is used in a fixed position compared to the prostate [2,3]. Also, the detection of whole implant displacement is challenging due to a lack of proper anatomical referencing.To address these limitations, we evaluated a single reference sensor solution enabling both motion compensation during EMT acquisition and inter-fraction implant displacement detection. Material/Methods: Paired planning CT images from 13 HDR-BT patients (single implant, 2 × 13.5 Gy in one day) were used to assess intra-patient motion and deformation. Differences in relative positions of potential internal and external reference sensor locations were quantified. Two external (on patient, stable in respect to the pelvic bone; on implantation template) and seven internal reference sensor positions along the Foley catheter were evaluated.Based on this analysis, a prototype was constructed with a reference sensor integrated at the optimal position within a standard two-way Foley catheter. This was evaluated in a dynamically moving prostate phantom with 6 implanted 6F needles (Figure 2.a). EMT data were acquired using an afterloader prototype with an EMT sensor integrated in the check-cable (Elekta AB, Stockholm, Sweden), stopping at 21 equidistant positions within each of the implanted needles. Post- processing and motion compensation were applied as described earlier [2,3]. Periodic phantom translations (range: 1–10 mm, frequency: 0.2–1.0 Hz) were applied during EMT measurements. Euclidean distance between the dwell position deviations in the static and in the moving setup with motion compensation was calculated. Also, EMT measurements were performed
Conclusion: Overall this project trained and evaluated three deep learning auto-segmentation models, with U-Net determined to be most accurate. The model generated contour dose variations were evaluated to be within the current inherent clinical variations for all ROIs other than Rectum, and to have the potential to realise significant time savings. Keywords: Auto-segmentation, MRI References: [1] Yin XX, Sun L, Fu Y, Lu R, Zhang Y. U-Net-Based Medical Image Segmentation. J Healthc Eng2022;2022.[2] Yang R, Yu J, Yin J, Liu K, Xu S. An FA- SegNet Image Segmentation Model Based on Fuzzy Attention and Its Application in Cardiac MRI Segmentation. International Journal of Computational Intelligence Systems 2022;15(1).[3] Zhao H, Shi J, Qi X, Wang X, Jia J. Pyramid Scene Parsing Network. 2016; Available from: http://arxiv.org/abs/1612.01105[4] Hellebust TP, Tanderup K, Lervåg C, Fidarova E, Berger D, Malinen E, et al. Dosimetric impact of interobserver variability in MRI-based delineation for cervical cancer brachytherapy. Radiotherapy and Oncology 2013;107(1):13–9.
Made with FlippingBook - Share PDF online