A neural network‐based 2D/3D image registration quality evaluator for pediatric patient setup in external beam radiotherapy

• DOI: 10.1120/jacmp.v17i1.5235
• Publication Date: 2017-02-28T00:23:13Z
• Authors: Jian Wu; Zhong Su; Zuofeng Li
• Keywords: Image registration; External beam radiotherapy

Our purpose was to develop a neural network‐based registration quality evaluator (RQE) that can improve the 2D/3D image robustness for pediatric patient setup in external beam radiotherapy. Orthogonal daily X‐ray images of six patients with brain tumors receiving proton therapy treatments were retrospectively registered their treatment planning computed tomography (CT) images. A pattern classifier used determine whether solution successful based on geometric features similarity measure values near point‐of‐solution. Supervised training and test datasets generated by rigidly registering pair orthogonal CT. The best each task selected from 50 optimizing attempts differed only randomly initial transformation parameters. distance individual normalized parametrical space compared user‐defined error tolerance acceptable. supervised then train RQE. Performance RQE evaluated using dataset consisting results not training. integrated our in‐house system its performance same dataset. With an optimized sampling step size (i.e., 5 mm) feature space, has sensitivity specificity ranges 0.865–0.964 0.797–0.990, respectively, when detect mean voxel displacement (MVD) greater than 1 mm. trial‐to‐acceptance ratio system, all patients, is equal 1.48. final acceptance 92.4%. proposed potentially be rigid overall rejecting unsuccessful solutions. patient‐specific, so single constructed particular application (e.g., acquired anatomical site). Implementation clinically feasible. PACS numbers: 87.57.nj, 87.85.dq, 87.55.Qr