Program Information
3D Markerless Registration of Lung Based On Coherent Point Drift: Application in Image Guided Radiotherapy
J Nasehi Tehrani1*, J Wang1 , X Guo2 , Y Yang3 , (1) UT Southwestern Medical Center, Dallas, TX, (2) University of Texas at Dallas, Richardson, Texas, (3) The University of New Mexico, New Mexico, NM
Presentations
SU-C-18A-4 Sunday 1:00PM - 1:55PM Room: 18APurpose: This study evaluated a new probabilistic non-rigid registration method called coherent point drift for real time 3D markerless registration of the lung motion during radiotherapy.
Method: 4DCT image datasets Dir-lab (www.dir-lab.com) have been used for creating 3D boundary element model of the lungs. For the first step, the 3D surface of the lungs in respiration phases T0 and T50 were segmented and divided into a finite number of linear triangular elements. Each triangle is a two dimensional object which has three vertices (each vertex has three degree of freedom). One of the main features of the lungs motion is velocity coherence so the vertices that creating the mesh of the lungs should also have features and degree of freedom of lung structure. This means that the vertices close to each other tend to move coherently. In the next step, we implemented a probabilistic non-rigid registration method called coherent point drift to calculate nonlinear displacement of vertices between different expiratory phases.
Results: The method has been applied to images of 10-patients in Dir-lab dataset. The normal distribution of vertices to the origin for each expiratory stage were calculated. The results shows that the maximum error of registration between different expiratory phases is less than 0.4 mm
(0.38 SI, 0.33 mm AP, 0.29 mm RL direction). This method is a reliable method for calculating the vector of displacement, and the degrees of freedom (DOFs) of lung structure in radiotherapy.
Conclusions: We evaluated a new 3D registration method for distribution set of vertices inside lungs mesh. In this technique, lungs motion considering velocity coherence are inserted as a penalty in regularization function. The results indicate that high registration accuracy is achievable with CPD. This method is helpful for calculating of displacement vector and analyzing possible physiological and anatomical changes during treatment.
Contact Email: