Quantitative evaluation of rigid and elastic registrations for abdominal perfusion imaging with X-ray computed tomography |
| |
Authors: | M-E Cohen C Pellot-Barakat J-M Tacchella M Lefort A De Cesare J Lebenberg N Souedet O Lucidarme T Delzescaux F Frouin |
| |
Institution: | 1. Laboratoire d’Imagerie Fonctionnelle, UMR_S 678 Inserm, université Pierre-et-Marie-Curie, CHU Pitié-Salpêtrière, 91, boulevard de l’Hôpital, 75013 Paris, France;2. Priam, ESME-Sudria, 38, rue Molière, 94200 Ivry-sur-Seine, France;3. MIRCen, I2BM, DSV, CEA, 18, route du Panorama, 92260 Fontenay-aux-Roses, France;4. Service de radiologie polyvalente et oncologique, CHU Pitié-Salpêtrière, AP–HP, 47-83 boulevard de l’Hôpital, 75013 Paris, France |
| |
Abstract: | ObjectivesDynamic X-ray computed tomography with contrast agent injection allows quantifying the tumor response to an antiangiogenic treatment. For hepatic tumors, it is crucial to correct for respiratory movements to ensure a good quantification of perfusion parameters.Material and MethodsTwo registration models were tested on eight dynamic exams: 1) rigid model and 2) elastic model based on Free Form Deformation. Each exam included about 50 volumes acquired under free-breathing conditions. The registration methods were quantitatively and automatically evaluated by defining a global score. This score was based on a Dice index and measured the superimposition rate between corresponding bony structures of different frames of the temporal exam.ResultsBefore registration, the mean value of the score was 0.78. After rigid registration, these values were globally lower with an average of 0.73. After elastic registration, a mean score of 0.90 was reached which was much higher than before registration.ConclusionThis automatic evaluation showed its interest for controlling the quality of dynamic CT registrations. It also helped proving that elastic methods are much more efficient than rigid methods for registering abdominal dynamic volumes, when considering the whole field of view. |
| |
Keywords: | |
本文献已被 ScienceDirect 等数据库收录! |
|