Close-to-actual delineation method in SPECT images

Identifying the border of a human´s organ in medical images, especially in SPECT (Single Photon Emission Tomography) images, is a very difficult task due to the high diffusion of radiopharmaceuticals around the boundaries. Many segmentation techniques have been proposed to segment and find the volume of internal organ from SPECT images. Results are usually compared with those from manual segmentation by experts. Nevertheless, experts´ manual segmentation results can easily be inaccurate because of the scattered boundaries in the SPECT images. This paper proposes a segmentation method to get close-to-actual boundaries in SPECT images. Geometric-shape phantoms are used for verifying the close-to-actual edges via volume calculations. Results from the proposed method are compared with those from manual segmentation method.