Abstract:
Targets and organs at risk (OARs) delineation is an essential process in radiation therapy. These structures provide information for treatment planning. The delineation process requires a lot of information using multiple imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET) and so forth. Due to difference in acquisition time and machines, image registration is required to map coordination between each imaging modality. This study investigated the performance of image registration methods for the structure delineation in head and neck cancer. This study recruited 22 treatment plans of cancer that were used for treatment. The structures from simulated computed tomography (CTsim) were used as the ground-truth for this study. The latest diagnostic computed tomography (CTdiag) of these selected plans was enrolled and imported into third-party delineation software to create the structures. The performance of image registration methods both rigid image registration (RIR) and deformable image registration (DIR) for delineation was measured using the Dice Similarity Coefficient (DSC) and Hausdorff distance (HD). The comparison involved not only a different image registration method but also a different treatment region (nasopharynx and oropharynx) This study found higher efficiency in structure delineation using DIR than RIR both in DSC and HD. The statistic showed no significant difference in the value of HD for the target and the larynx. In nasopharynx, a significant difference was observed in all structures except the right eye in DSC, whereas HD revealed a significant difference in the spinal cord, right parotid and left eye. In the oropharynx, no significant difference was found in the delineation of the larynx in DSC, and the target and right parotid in HD. This study confirmed that DIR was effective for delineating OARs with the DSC value showing over 0.7, whereas it was providing a useful guideline in this regard."