SeminarsThe Introduction of Reconstruction and Segmentation on Medical Images: Applied for Positron Emission Tomography
63
reads
Reconstruction and segmentation of medical images are always an essential and important goal for the clinical and preclinical applications. The positron emission tomography (PET) typically generates functional metabolic distribution of human for the injection of isotope tracers. The magnetic resonance imaging (MRI) provides super detail resolution of 1H density of human body. The computed tomography (CT) is often used to image anatomical disorder, such as lung cancer, myocardial artery disease (CAD), and calcium or stone in tissues. However, it has to take long scan time to acquire high resolution images for diagnosis by applied MRI or PET. On the other hand, the volumes of tumor or lesion are important information for the plan of therapy treatment. In this talk, the joint Poisson model was applied to reconstruct PET images in short times. Meanwhile, the Gaussian mixture model combined with kernel density estimation was used to segment cerebral area of rat brain PET images. The proposed reconstructed method provides less noise and high image quality of PET. The presented segmented algorithm generated clearer and more detailed structures of an FDG accumulation location in the cerebral cortex than those by the K-Means method. (Keywords: PET, Cerebral cortex, K-Means, Hybrid Gaussian mixture model, Kernel density estimation )
reads
Tai-Been Chen
2013-11-01
12:30:00 - 14:30:00
103 , Mathematics Research Center Building (ori. New Math. Bldg.)
Reconstruction and segmentation of medical images are always an essential and important goal for the clinical and preclinical applications. The positron emission tomography (PET) typically generates functional metabolic distribution of human for the injection of isotope tracers. The magnetic resonance imaging (MRI) provides super detail resolution of 1H density of human body. The computed tomography (CT) is often used to image anatomical disorder, such as lung cancer, myocardial artery disease (CAD), and calcium or stone in tissues. However, it has to take long scan time to acquire high resolution images for diagnosis by applied MRI or PET. On the other hand, the volumes of tumor or lesion are important information for the plan of therapy treatment. In this talk, the joint Poisson model was applied to reconstruct PET images in short times. Meanwhile, the Gaussian mixture model combined with kernel density estimation was used to segment cerebral area of rat brain PET images. The proposed reconstructed method provides less noise and high image quality of PET. The presented segmented algorithm generated clearer and more detailed structures of an FDG accumulation location in the cerebral cortex than those by the K-Means method. (Keywords: PET, Cerebral cortex, K-Means, Hybrid Gaussian mixture model, Kernel density estimation )