Premium
Fully automatic deformable registration of pretreatment MRI / CT for image‐guided prostate radiotherapy planning
Author(s) -
Hamdan Iyas,
Bert Julien,
Rest Catherine Cheze Le,
Tasu Jean Pierre,
Boussion Nicolas,
Valeri Antoine,
Dardenne Guillaume,
Visvikis Dimitris
Publication year - 2017
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 - 10.1002/mp.12629
Subject(s) - image registration , computer science , hausdorff distance , artificial intelligence , radiation treatment planning , computer vision , mutual information , robustness (evolution) , medical imaging , radiation therapy , medicine , radiology , image (mathematics) , biochemistry , chemistry , gene
Purpose In prostate radiotherapy, dose distribution may be calculated on CT images, while the MRI can be used to enhance soft tissue visualization. Therefore, a registration between MR and CT images could improve the overall treatment planning process, by improving visualization with a demonstrated interobserver delineation variability when segmenting the prostate, which in turn can lead to a more precise planning. This registration must compensate for prostate deformations caused by changes in size and form between the acquisitions of both modalities. Methods We present a fully automatic MRI / CT nonrigid registration method for prostate radiotherapy treatment planning. The proposed registration methodology is a two‐step registration process involving both a rigid and a nonrigid registration step. The registration is constrained to volumes of interest in order to improve robustness and computational efficiency. The method is based on the maximization of the mutual information in combination with a deformation field parameterized by cubic B‐Splines. Results The proposed method was validated on eight clinical patient datasets. Quantitative evaluation, using Hausdorff distance between prostate volumes in both images, indicated that the overall registration errors is 1.6 ± 0.2 mm, with a maximum error of less than 2.3 mm, for all patient datasets considered in this study. Conclusions The proposed approach provides a promising solution for an effective and accurate prostate radiotherapy treatment planning since it satisfies the desired clinical accuracy.