形状解析のための画像処理：mathematical morphology に基づく画像コンポーネントの抽出

抄録

Image processing is a crucial step in the quantitative interpretation of biomedical images acquired from the various biomedical modalities, such as mammography, X-ray computed tomography, magnetic resonance imaging and light and electron microscopy. One of the most important purposes of the image processing is to derive meaningful information for utilizing the shape analysis.The process of shape analysis consists of two main steps: (1) the extraction of image components of the target (e.g., area, boundary, network pattern and skeleton), (2) the description of the shape features (e.g., size, perimeter, circularity and compactness). In this study, a target region extraction approach based on a new type of mathematical morphology is introduced. Mathematical morphology is a nonlinear image processing method based on the set theory and is useful for the extraction of the image components from an image. It can be used as a fundamental tool to analyze the biomedical images. However, conventional morphological processing has a common drawback that a unique type of artifact might appear in the processed image, depending on a fixed scanning direction of the structuring element (SE). Since biomedical images consist of delicate shape features which have directional variations, the artifact is caused by the insufficient morphological processing with the limitation of the SE scanning direction.To overcome this problem, an extension of conventional morphological operation has been proposed. The algorithm of this new type morphological operation uses a single SE to a series of rotations of the original image. Since the algorithm enables isotropic processing, it causes a considerable reduction in this type of artifact. In this paper, the new algorithm of the image processing method is applied to extraction of nuclei regions in histological sections of Medaka testis image for quantification of the number and the size of nuclei.