The Research Of PET Image Denoising Based On Kernel ICA

With the development of medical imaging and computer-aided technology, medical image processing technology is currently one of the fastest growing technologies. As kinds of new medical imaging methods are used in the clinical diagnosis, the bodies of the lesion can be observed more directly, clearly and with higher confirmation rate, thereby great progress has been made in medical diagnosis and techniques of treatment. Because Positron Emission Computed Tomography (PET) can provide information of body functions, PET are widely used in the clinical practices. The image processing of PET becomes the Hotspot, so this discertation mainly studied the disposing of PET image.The discertation firstly outlines the historical development of medical imaging, medical image processing technology and filtering algorithm commonly used in medical image.Secendly, this discertation introduces the development of PET technology, PET principles, the PET/CT technology evolved by PET and development prospects of PET/CT.Third, this discertation emphasized the development history, basic principle, constraints and the research status of Independent Component Analysis (ICA). This discertation also studied Fixed-Point Algorithm and Kernel Independent Component Analysis.This discertation proposed a method which applied the Kernel ICA to PET image denoising processing. We pre-process the experimental data of the PET image using Principal Component Analysis (PCA). After we denoise the image data got from the pre-prossing using the Kernel ICA. We reconstruct PET images to obtain better effects of the PET images.Finally experimental results show that comparing with the traditional ICA de-noising algorithm in medical image processing, the de-noising algorithm used in this discertation is more suitable for PET image processing. The Kernel ICA can highlights effects of the image by different Kernel functions. So the Kernel ICA is a very practical and efficient algorithm in the PET image de-noising processing.