Evaluation of Image Processing Methods for ANA Immunofluorescence Images

• Main Authors: Chen-Yi Lee, 李貞儀
• Other Authors: Jinn-Yi Yeh
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/584962

碩士 === 國立嘉義大學 === 資訊管理學系碩士班 === 96 === The expert of immunology and rheumatology department inspects autoimmunedisease by recognizing antinuclear autoantibodies(ANA) patterns. HEp-2 cells are used for identification of ANA. This technique can recognize over thirty different nuclear and cytoplasmic patterns, which are given by upwards of 100 different autoantibodies. Therefore, physician diagnoses patients’ disease by inspecting HEp-2 cells. So far, identification of ANA is completed by inspecting the slides with the help of a fluorescent microscope. This manual procedure requires highly specialized technicians and consuming time. There are many image processing and data mining methods for classifying ANA images automatically. This research attempts to find the cascaded method having highest accuracy from different image processing methods. These methods include edge detection, feature selection, and classification. Initially, we acquire ANA images and transfer them to be gray-level images. Edge detection methods are then used to locate cells. The next step is to find important features. Finally, we use classification methods to predict images’ classes. The result shows that the optimal cascaded method is the combination of Canny (edge detection method), support vector machine (feature selection method), and support vector machine (classification method). The accuracy rate is about 97.00%. Therefore, we suggest that physician can identify ANA by using this cascaded method to diagnose disease in practice.