Image Enhancement and Border Detection of Parotid Sjogren''s Syndrome in Magnetic Resonance Imaging

• Main Authors: Bo-syun Chen, 陳柏勳
• Other Authors: Ja-chih Fu
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/61752847822079710467

碩士 === 國立雲林科技大學 === 工業工程與管理研究所碩士班 === 97 === Within the diagnosis history of Parotid Sjögren''s Syndrome, ultrasonic has often been used to observe the extension of salivary duct to determine the degree of patients’ disease. There are diverse molds and feature extraction ways to examine Parotid Sjögren''s Syndrome. Recently, there have been several researches which determined the degree of patients suffering from Parotid Sjögren''s Syndrome by the image of T1-Weighted and T2-Weighted, what degrees of intact lobule area lessen and what degrees of fat area increase. MRI (Magnetic Resonance Imaging) or Nuclear Magnetic Resonance is one of the crucial ways to acquire medical image. Through computer images, MRI is capable of presenting human organs, tissue structures, and the dimensional sections of nidus. MRI examining neither invades nor causes damage to human bodies; it is free from the injury of ray and its effectiveness is high, time for examining is short, it can not only have several dimensional scan but provides 3D special images. MRI can be an indispensable diagnostic tool. However, current MRI of Parotid Sjögren''s Syndrome lacks disposal of entire image intensity, which leads to inferior contrast and vagueness of images. This research aims at the MRI of Salivary Gland to establish an algorithm, which can strengthen the contours of Salivary Gland. Through Principal Component Analysis, the algorithm combines theories such as Weighted Histogram Separation with Independent Component Analysis, deriving strengthened contours of the image of Salivary Gland. Inputting the original images and the strengthened images from Weighted Histogram Separation, we get several image sets inclusive of independent component analysis of intensified first image, independent component analysis of intensified second image, main component analysis of intensified first image, main component analysis of intensified second image, main component analysis of intensified third image, main component analysis of intensified forth image. According to PSNR, the quality of main component analysis of intensified first image (average: 22.146, standard deviation:3.1413 ) is superior to the quality of the image of Weighted Histogram Separation (average: 18.828, standard deviation:1.9898 ). Two of them both have eminent diversity; therefore, main component analysis of intensified first image can be the most representative input image. Through the 3D-image of parotid, we can find abnormal volume of salivary gland would wither or some protuberances would appear on the smooth, and this visual effect can be served as references for doctors to diagnose Parotid Sjögren''s Syndrome. This research uses Skeletonization algorithm to calculate the volume of Salivary Duct and utilizes the relationship among numbers of average pixel and the value of pixels and the relationship between the volume and length of ducts to compute the area inaccuracy between all average sections and two average sections. (The rate of inaccuracy is 3.00%~7.70%) In the future research of Skeletonization, researchers can take advantage of this tool to get and measure features in the vessels.