J. Noyola-Martinez, T. Guerrero, Y. Borghero, K. Sanders, C. Stevens, M. Guerrero, and R. Guerra (USA)
Respiratory tumor motion, mutual information, computed tomography, treatment planning.
Purpose: To develop and perform statistical evaluation of a technique to map physician delineated gross tumor volumes (GTV) across multiple CT images representing phases of the respiratory cycle. Materials & Methods: A series of six radiation treatment planning exhalation and inhalation breath-hold CT images (eBH-CT & iBH-CT) from lung cancer patients were selected. The GTV was delineated on each by nine thoracic radiation oncology physicians. The eBH-CT and iBH-CT images were registered using normalized mutual information for a sub-region surrounding the tumor. The eBH-CT GTV was mapped to the iBH-CT. The mapped GTV was compared with the physician delineated iBH-CT GTV using the Dice similarity coefficient (DSC), a spatial overlap index. Results: The tumor centrioid respiratory induced motion for six cases measured from 0.065 cm to 2.96 cm. The tumor volumes ranged from 7.43 cm3 to 493.3 cm3 . For the unmapped GTV pairs the DSC ranged from 0.0 to 0.91 with mean 0.75, and for the mapped GTV the range was 0.22 to 0.91 with mean 0.81, which represents a significant (p < 0.001) increase. In comparison the DSC between physicians GTVs (inter-observer) within each case ranged from 0.37 to 1.0 (mean 0.75). Conclusions: Rigid body registration of sub-image regions surrounding the tumor is a feasible method to map tumor regions across CT images that represent various phases of the respiratory cycle.
Important Links:
Go Back
