Free Form Deformable Model Based On Multi-Resolution B-Spline Grid for Registration of 4D CT Images of Hepatocellular Carcinoma Patients
H Wang1*, D Li2, H Wang3, Y Yin4, G Gong5, (1)Shandong Unversity ,Jinan,Shandong Province,(2) Shandong Normal University, JI nan, Shandong Province, (3) Shandong Unversity, Jinan,Shandong Province,(4) ,jinan, ,(5) Shandong Cancer Hospital, Jinan, Shandong ProvinceSU-E-J-86 Sunday 3:00:00 PM - 6:00:00 PM Room: Exhibit Hall
Purpose: The breathing motion has a significant influence to radiation therapy of hepatocellular carcinoma patients, and the motions are irregular. We used deformable registration of four-dimensional computed tomography (4D CT) data to eliminate breathing motion artifact to truly track the tumors' positions and shapes, and guide medical physicists to delineate the more accurate tumor targets for treatment planning.
Methods: We collected 4D CT data of 8 hepatocellular carcinoma patients, 4D-CT images were segmented into 10 serious CT images according the respiratory phase, and named CT0,CT10'¦CT90, and CT0 was at end inspiration , CT50 was at end expiration. We proposed the free form deformable (FFD) model based on multi-resolution B-spline grid registration method to the 4D CT data during a whole breathing cycle, and we used L-BFGS to optimize system and multi-resolution to pick up the registration speed and accuracy. In particular we defined the gross target volume of HCC with the help of iodipin on CT images which can verify the registration performance. We chose end-exhale CT as the reference image.
Results: We compared the floating images, the subtractions of floating images and reference image, the iodipin landmarks' positions and the multi-informations between floatings image and reference image before and after registered to verify the registration performance. From the registered images, we can track the 'trajectory' of the tumors during a whole breathing cycle, and the average multi-information improved 3.5% after registration between the two extreme phases.
Conclusions: We can use this registration method to register the 4D CT data and use iodipin on CT images to define gross target volume to guide medical physicists delineating the more accurate tumor targets.