Premium
Registration of clinical volumes to beams‐eye‐view images for real‐time tracking
Author(s) -
Bryant Jonathan H.,
Rottmann Joerg,
Lewis John H.,
Mishra Pankaj,
Keall Paul J.,
Berbeco Ross I.
Publication year - 2014
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4900603
Subject(s) - image guided radiation therapy , image registration , tracking (education) , hounsfield scale , computer vision , artificial intelligence , medical imaging , cone beam computed tomography , computer science , nuclear medicine , radiography , volume (thermodynamics) , medicine , computed tomography , radiology , image (mathematics) , physics , psychology , pedagogy , quantum mechanics
Purpose: The authors combine the registration of 2D beam's eye view (BEV) images and 3D planning computed tomography (CT) images, with relative, markerless tumor tracking to provide automatic absolute tracking of physician defined volumes such as the gross tumor volume (GTV). Methods: During treatment of lung SBRT cases, BEV images were continuously acquired with an electronic portal imaging device (EPID) operating in cine mode. For absolute registration of physician‐defined volumes, an intensity based 2D/3D registration to the planning CT was performed using the end‐of‐exhale (EoE) phase of the four dimensional computed tomography (4DCT). The volume was converted from Hounsfield units into electron density by a calibration curve and digitally reconstructed radiographs (DRRs) were generated for each beam geometry. Using normalized cross correlation between the DRR and an EoE BEV image, the best in‐plane rigid transformation was found. The transformation was applied to physician‐defined contours in the planning CT, mapping them into the EPID image domain. A robust multiregion method of relative markerless lung tumor tracking quantified deviations from the EoE position. Results: The success of 2D/3D registration was demonstrated at the EoE breathing phase. By registering at this phase and then employing a separate technique for relative tracking, the authors are able to successfully track target volumes in the BEV images throughout the entire treatment delivery. Conclusions: Through the combination of EPID/4DCT registration and relative tracking, a necessary step toward the clinical implementation of BEV tracking has been completed. The knowledge of tumor volumes relative to the treatment field is important for future applications like real‐time motion management, adaptive radiotherapy, and delivered dose calculations.