Study on Image Processing of Capillaries Using Microscope: Initial Considerations

Hang Nguyen Thi Phuong, Hieyong Jeong, Choonsung Shin

Abstract: Routine nailfold capillaroscopy is essential for early diagnosis of rheumatic disease, but clinicians struggle to evaluate the same capillaries in every examination because the capillaries are similar in shape; therefore, difficult to differentiate. Here, we propose a method that will enable us to compare previous and real-time images captured by a microscope with low- and high-power objective lenses. Particularly, we would like to look into the method for feature extraction such as template matching, SIFT (Scale Invariant Feature Transform), and so on. After selecting a target nailfold capillary using the low-power objective lens, we assessed the capillary with a high-power objective lens. The captured image of the target capillary was compared in real-time with the previously captured image, irrespective of magnification, using template matching. Our proposed algorithm detected allowable similarity for assessing the same capillary, which was helpful in evaluating the structural abnormalities of the entire area in addition to just the magnified area.