User: Guest  Login
Title:

An X-ray and computed tomography based registration method for patient to beam alignment designed for high reliability despite of degraded X-ray image quality

Document type:
Konferenzbeitrag
Author(s):
Selby, B. P.; Walter, S.; Sakas, G.; Groch, W.-D.; Stilla, U.
Abstract:
Background: To benefit from particle beam cancer treatment accuratetumor to beam alignment is crucial. In image guided radiotherapy(IGRT) this is done by comparison of reconstructed radiographs (DRRs)from pre-treatment CTs and stereoscopic intra-treatment digital radiography(DR) images. Mutual Information (MI) based similarity measures (SM)have been found suitable for geometric registration of these images.But there are several factors influencing DR quality that can degraderegistration performance, as vignetting, low contrast e.g. in pelvicimages, noise and non-rigid transformations through movement of bodyparts (e.g. femurs). To overcome those obstacles, we present a novelmethod, improving alignment reliability and reducing computationtime.Material and Methods: With low quality DRs, automatic algorithms oftenfail to perform registration to DRRs, thus manual registration isoften still possible. We therefore propose a full automatic methodthat in a first step omits dispensable information by extractinggradients of the DRs and the DRRs. Because DRs can contain a highamount of noise, we use large (˜ 80 pixels) filter masks to extractgradients from a large neighborhood. To decrease computation timewe use filter operations without different weights for neighboringpixels and work on integral images to summarize over large portionsof data with only a few operations. The image gradients are smoothedto remove noise and to increase their range of influence and theprobability to find extrema in the SM during maximization. Informationin the neighborhood of a gradient helps to decide which gradientsto use and which to omit. Finally, the SM is computed by an MI basedapproach modified through weights from gradient intensity differencesto improve stability. Iterative maximization of the SM for 2 pairsof images gives the rigid transformation of the patient body in 6degrees of freedom (DOF). Results: We compared common MI (AP1) with our new approach (AP2) usingA) high quality DRs as well as B) low quality images and C) DRs,degraded, so that even manual registration was nearly impossible.For cases A, both approaches succeed, AP1 being slightly more accurate.For cases B, the advantages of the new approach became clear. AP1failed for some datasets, but AP2 still was able to register theimages correctly. For cases C, mostly pelvic images, AP1 failed foralmost every dataset, but AP2 performed well.Conclusion: We presented a new approach for 6 DOF alignment, ableto overcome problems resulting from low quality DRs. For images e.g.of the head & neck, where bony structures become clear, our approachperforms as well as a common MI based registration, but it becomesmuch more reliable for images of the pelvic or thorax region, wherelow contrast and irregular dose distribution degrade the DRs. Comparedto common MI, our approach reduced the computation time by factor2 as global maxima of the SM can be found within less iterations.
Keywords:
Infrared images, image sequences, GPS, inertial navigation, 3D city models, edge matching
Book / Congress title:
Proceedings of the 48th Conference of the Particle Therapy Co-Operative Group
Year:
2009
Pages:
128--129
WWW:
http://www.pf.bv.tum.de/stilla/pub.html
 BibTeX