ESTRO 2022 - Abstract Book

S175

Abstract book

ESTRO 2022

Conclusion The proposed Deep Learning segmentation algorithm performed generally very well showing similar performance as the experts for all structures except the optical tracts when comparing DSCs. This discrepancy may be attributed to the small volume of these structures where the DSC is known to be especially sensitive to small variations in delineations. The expert segmentations with large deviation from the general trend of the expert segmentations (HD95 well beyond the interquartile range) were visually inspected and was found not to adhere to the DNOG guidelines. The auto segmentations were on the other hand in good agreements with the general expert trend and a potential added benefit of auto segmentation could be outlier detection of subpar segmentations for example in retrospective datasets. To conclude, this deep learning network is of such high quality as to be used in clinical workflow and shows potential as a tool for delineation outlier detection.

MO-0218 A likelihood-based particle imaging filter using prior information

R. Fullarton 1 , L. Volz 2,3,4 , N. Dikaios 5,6 , R. Schulte 7 , G. Royle 8 , P. Evans 9,10 , J. Seco 11,12 , C. Collins Fekete 13

1 University College London, Department of Medical Physics and Biomedical Engineering, London, United Kingdom; 2 Heidelberg University, Department of Physics and Astronomy, Heidelberg, Germany; 3 German Cancer Research Centre, Biomedical Physics in Radiation Oncology, Heidelberg, Germany; 4 GSI Helmholtz Centre for Heavy Ion Research GmbH, Biophysics, Darmstadt, Germany; 5 University of Surrey, Centre for Vision Speech and Signal Processing, Guilford, United Kingdom; 6 Academy of Athens, Mathematics Research Center, Athens, Greece; 7 Loma Linda University, Basic Sciences, Loma Linda, USA; 8 University College London, Department of Medical Physics and Biomedical Engineering, London, United Kingdom; 9 University of Surrey, Centre for Vision Speech and Signal Processing, Guilford, United Kingdom; 10 National Physical Laboratory, Chemical, Medical and Environmental Science, Teddington, United Kingdom; 11 German Cancer Research Centre, BioMedical Physics in Radiation Oncology, Heidelberg, Germany; 12 Heidelberg University, Department of Physics and Astronomy, Heidelberg, Germany; 13 University College London, Department of Medical Physics and Biomedical Engineering , London, United Kingdom Purpose or Objective Particle imaging can increase precision in proton and ion therapy. Interactions with nuclei in the imaged object result in image noise and reduced image quality, especially for multi-nucleon ions that may fragment. This work proposes a filter based on the physics of electromagnetic interactions to identify and remove ions, that have undergone nuclear interaction, and hence contribute to image noise. Materials and Methods The filter combines a prior reconstruction with scattering and straggling theory to determine the likelihood that a particle only interacts electromagnetically (primary). A threshold ( P t ) is then set to reject those particles with a low likelihood (secondary). The filter performance was compared with the state-of-the-art 3 σ filter which removes particles based on the water equivalent thickness (WET) and angular distributions per pixel. We reconstructed proton and helium radiographs from simulated data of the XCAT thorax phantom. Experimental proton and helium CT scans of a Catphan 404 Sensitometry module (The Phantom Laboratory, Salem, NY, USA) were also evaluated. The radiographs and the tomographic reconstructions were evaluated based on intra-pixel noise.