Summary: | 碩士 === 東海大學 === 資訊工程學系 === 98 === Indirect immunofluorescence (IIF) with HEp-2 cells has been used to detect antinuclear auto-antibodies (ANA) for diagnosing systemic autoimmune diseases. An automatic inspection system for ANA testing can be partitioned into HEp-2 cell detection, fluorescence pattern classification and computer aided diagnosis phases. The aim of this study is to develop an automatic classification scheme to identify the fluorescence patterns of HEp-2 cell in IIF images. The proposed method firstly utilizes a preprocessing procedure to reduce noise within IIF images. The feature extraction step calculates 14 practical textural and statistic features from fluorescence patterns. Then the proposed classification method performs an unsupervised neural network, the self-organizing map (SOM), to identify each fluorescence cell. This study evaluates 1020 autoantibody fluorescence patterns which were manually divided into six primary patterns, i.e. diffuse, peripheral, coarse speckled, fine speckled, discrete speckled and nucleolar patterns. The k-fold cross-validation method is used to evaluate the classification performance. From the simulations, the average accuracy using the proposed approach for identifying ANA patterns is 92.4%. The result shows that the proposed approach to identify autoantibody fluorescence patterns with a high accuracy and is therefore clinically useful to provide a second opinion for diagnosing systemic autoimmune diseases.
|