Automatic segmentation of structures in CT images for head and neck intensity-modulated radiation therapy

Abstract

Cancers in the head and neck region account for approximately 3 percent of all cancers in the United States, as it is reported by the American Cancer Society. Depending on the location and stage of the cancers, surgery, chemotherapy and radiation therapy are considered the major treatment options. Since the past decade intensity-modulated radiation therapy (IMRT) has become the state of the art in head and neck radiation therapy, segmentation of head and neck structures to be treated, including the level II, III, and IV lymph nodes, as well as structures to be spared, including the thyroid glands and the parotid glands, from the diagnostic computed tomography (CT) images is of great importance for treatment planning. In order to reduce the time required to manually segment these structures, a set of innovative approaches are proposed to automate the process: An active shape model (ASM) is constructed to segment the level II, III, and IV lymph node regions; a multiple-atlas-based approach is implemented to segment the thyroid gland; a constrained ASM with landmark uncertainty is used to segment the bilateral parotid glands. Both qualitative and quantitative results have shown that the automatically generated segmentations are of high precision, and the proposed approaches have the potential to reduce the delineation efforts required for clinical IMRT treatment planning.