S2075
Physics - Image acquisition and processing
ESTRO 2026
Poster Discussion 3535
maps), as shown in Fig 1. The selected samples were strategically distributed across the low-risk CTV, but outside the high-risk CTV, to maximize the expected information gain while avoiding high-risk or off-limit regions for biopsy sampling as defined by proximity to the critical organs (i.e., brainstem, hippocampus, optic nerves, chiasm, and pituitary gland). Assuming the probability of a positive biopsy at each voxel follows a Bernoulli distribution, Tab 1 reports the rate of low- risk CTV rejection as a function of biopsy results and rejection criteria, suggesting a high rate of potential biopsy-driven reduction of radiation target volume, especially for K equal to 1 or 2.
Motion detection and mitigation in consecutive dual-energy CT for robust motion-aware treatment planning Sebastian Baum 1,2 , Annette Schwarz 1,2 , Patrick Wohlfahrt 1 1 Cancer Therapy Imaging, Siemens Healthineers, Forchheim, Germany. 2 Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany Purpose/Objective: Consecutive dual-energy CT (cDECT) enables spectral imaging on all CT scanners. While motion-induced anatomical changes during cDECT can lead to artifacts in DECT post-processing, its occurrence is relevant for radiotherapy planning.This study aims to develop an image-based framework for motion detection and mitigation to generate motion-artifact-free DECT images for both consecutive anatomical representations. Material/Methods: Forty thoracic and abdominal four-dimensional photon-counting CT (4DCT) scans were acquired on NAEOTOM Alpha.Peak (Siemens Healthineers). Motion was simulated by combining low- and high-energy images (L&H) from different breathing phases, covering variable motion extents from none (artifact- free reference, REF) to maximum inhale-exhale (artifact sample, ART) (Figure1A). While cDECT post- processing utilizes Demons registration for L&H alignment, residual motion-related artifacts are tracked with a refined motion mask originating from subtracting REF and ART electron-density images (DIF), and replaced with REF. Registration performance with varying pyramid levels (shallow S, deep D) was quantified using Dice similarity coefficient (Dice) on organ segmentations and mask coverage. Motion mask quality was assessed organ-specifically with structure similarity index and mean absolute error. The overall algorithm performance was evaluated by comparing exemplary photon and proton treatment plans applied to REF, as well as unprocessed original and motion-mitigated ART electron-density images.
Conclusion: This proof-of-principle study is the first to link biopsy sampling during surgery with downstream radiation planning through a mathematical framework, enabling neurosurgeons with a strategy to collect biopsy samples and verify the true extent of microscopic disease that is typically treated with high dose radiation without histologic verification. This information can ultimately inform radiation planning by excluding neighboring uninvolved brain tissue from the high-dose radiation volume. Keywords: Glioma, surgical biopsy, CTV
Made with FlippingBook - Share PDF online