MRI-only brain radiotherapy verification using cone beam computed tomography

Abstract

Introduction CBCT are frequently used for patient treatment verification. Hence, generated synthetic-CT (sCT) for MRI-only treatment planning workflow needs to be assessed for image guided radiation therapy (IGRT) using CBCT as well. Purpose For brain cancer patients, we investigated the effectiveness of using sCT that are generated from MRI instead of CT for CBCT-based IGRT. Material/methods Twelve patients with co-registered CT and conventional MRI were used to compute a library of patches. For three patients with two CBCT each, two sCT were generated from MRI using “non-local means” patch-based method (NMPBM) or “sparse” patch-based method (SPBM). CT and sCT alignments with CBCT were performed with MI-Powell. MI-Powell performs Powell optimization of mutual information between volumes and estimates alignment Euler angles and translation. Rigid transformations obtained for CT/CBCT registration were considered as the gold standard. Mean Target Registration Error (mTRE) obtained for sCT/CBCT alignment was computed. mTRE is the average distance between transformed CBCT voxels centers, within head region, by estimated and gold transformation respectively. Moreover, the average errors of displacement (Δ t ) and orientation (Δ θ ) from CT/CBCT registration transformation were computed. Results For sCT NMPBM /CBCT registration assessment, average mTRE = 1.36 ± 0.29 mm, Δ t = 0.92 ±0.2 mm and Δ θ = 1.02 ±0.34°. For sCT SPBM /CBCT registration assessment, average mTRE = 1.4 ± 0.3 mm, Δ t = 0.91 ± 0.21 mm and Δ θ = 1.08 ± 0.43°. Both NMPBM and SPBM were therefore accurate and gave an intermediate sCT suitable for CBCT-based IGRT. Conclusion Based on our results for brain site, CBCT can be used for patient position verification with sCT as a reference. Further work will be done for assessment in other clinical sites with challenging organ motion.