ESTRO 2025 - Abstract Book

S1774

Clinical – Upper GI

ESTRO 2025

Conclusion: Open-source algorithm performance is comparable to commercial algorithms, albeit with variation across upper abdominal OARs, and may be beneficial in treatment settings where a commercial algorithm is not economically viable. A key clinical finding is the potential disagreement between human and auto-contouring tools in the position of the duodenal/stomach junction. We highlight this as an important consideration when using auto-contouring tools, particularly if the duodenum or stomach is a dose-critical structure (e.g. for upper abdominal SBRT).

Keywords: AI, auto-contouring, organs at risk

1553

Digital Poster Does deep inspiration breath hold (DIBH) compress abdominal organs thus complicating radiotherapy for abdominal cancer? Sophia Liang 1 , Yilin Liu 1 , Chuan Zeng 1 , Xingyu Nie 2 , Ellen Yorke 1 , Wei Lu 1 , Guang Li 1 1 Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, USA. 2 Department of Radiology, Kentucky University College of Medicine, Lexington, USA Purpose/Objective: As deep inspiration breath hold (DIBH) pushes the diaphragm inferiorly compressing the organs in the abdomen and involving other abdominal muscles, it may reduce tumor-organ separation and anatomic reproducibility. We intend to test this hypothesis and compare organ deformation and changes from free breathing (FB) to DIBH. Material/Methods: We retrospectively analyzed the FB and two DIBH computed tomography (CT) images of 42 abdominal cancer patients acquired at simulation to determine organ compression and reproducibility of DIBH anatomy. A deep learning-based auto-segmentation plug-in program in MIM (version 7.2, MIM Maestro) was employed to contour normal structures in the abdomen and thorax. Based on the liver contour, relative volume changes from FB to DIBH1 and DIBH2 were calculated. Additionally, the bounding boxes of the liver were computed by identifying the maximum points in the superior-inferior, anterior-posterior, and left-right directions, and analyzed for their dimension changes in these directions. Further, internal fiducials or native landmarks, such as bifurcation points of the liver vessels and bilomas, were manually contoured in 18 patients, allowing regional compression analysis. Finally, deformable image registration was applied to assess tissue compression and expansion via the Jacobian determinant of the deformation vector field. Results: The distribution of liver volume (-2.0±11.7%) and superior-inferior dimension changes (-2.4±11.9mm) are diverse: 67% experience volume decrease while 33% have volume increase from FB to DIBH, using deep-learning-based auto-contour evaluations. This is consistent with diaphragm displacement (67% move 1.5-5.0cm and 33% move <1.5cm). Among the liver-compressed patients (67%), the superior portion of the liver above the mid-liver markers is compressed more than the inferior portion due to diaphragmatic pushing inferiorly, consistent with preliminary Jacobian analysis. In one patient, a large biloma is compressed from spherical in FB to oval in DIBH, with 25% shorter in the superior-inferior axis. For the liver-expanded patients (33%), increases in the inferior and/or anterior dimensions were observed owing to abdominal muscle involvement to provide more space. Marker positional uncertainty between DIBH1 and DIBH2 is 4.1±2.5mm, including both compression and expansion cases. Conclusion: This study demonstrates a broad distribution of liver compression or expansion in DIBH relative to FB: Liver compression may make organ sparing in DIBH more difficult and less manageable and abdominal muscle involvement may reduce DIBH setup reproducibility. Further investigations are granted to pursue shallow BH or