(Sunday, 7/29/2018) 3:00 PM - 6:00 PM
Room: Exhibit Hall
Purpose: Recently deformable image registration (DIR) playing important role in radiotherapy and Brachytherapy. Although there many advantages of using DIR especially in the planning of radiotherapy and Brachytherapy treatments but unfortunately a challenging task is choose a an accurate algorithm for specific procedure or even organs because of difficulties of verification process i.e to calculate the accuracy of DIR algorithms. The aim of this study was to evaluate a the accuracy of frequently used DIR algorithm called diffeomorphic demons using synthetic images from a virtual phantom.
Methods: For DIR process, synthetic reference(target)and wrapped(moving) images have been produced. Wrapped images were globally and locally deformed using radial basis function with different kernel. Thin Plate Spline and support radial basis functions kernels were utilized for global and local deformation respectively(figure1). DIR process has been done using the open-source 3D slicer toolkit applying diffeomorphic demons algorithm to unwrape the moving image sets with gradient type of 2 mm maximum step length. Then three regions of interests were delineated on the target and unwrapped images and the accuracy of registration was calculated using dice similarity coefficient and Hausdorff distance.
Results: The mean dice similarity coefficient between target image and unwrapped with global deformation images were found as 0.95. Also, the above coefficient between target image and unwrapped with local deformation were found as 0.99. Also the maximum Hausdorff distance was found as 1.93 and 1.85 mm for the unwrapped globally and locally deformed image sets respectively.
Conclusion: This study showed that applying DIR algorithm on a synthetic images sets from a virtual phantom can be considered as a valuable verification process to evaluate different DIR algorithm. Also our results showed that diffeomorphic demons algorithm is capable to unwrapped the globally or locally deformed images with reasonable accuracy.
Keywords
Not Applicable / None Entered.
Taxonomy
IM/TH- Image registration : General (Most aspects)
Contact Email
