Computer-aided Diagnosis For Immune Diseases In Antinuclear Antibodies Analysis

Medical imaging plays an important role in clinical diagnosis and treatment with the development of computer technology. More and more medical imaging modalities, such as radiography, magnetic resonance imaging, nuclear medicine, and photoacoustic imaging, have been applied to diagnose various diseases in the modern clinical diagnosis. Indirect immunofluorescence(IIF) based on HEp-2 cells substrate is the most commonly used staining method for antinuclear autoantibodies(ANA) associated with different types of autoimmune pathologies. These cells can reveal different staining patterns, e.g. coarse speckled, fine speckled, nucleolar and peripheral patterns.In this thesis, a new statistical-based algorithm, which applies the ideas of reduced binary pattern and multiple angles scheme, is proposed to extract image feature for staining pattern classification. The proposed statistical-based algorithm, multi-angle reduced binary pattern(RBP), is tested on the VGHTC dataset and sperm cell image database.We design an automatic system to identify the staining patterns based on block segmentation compared with the cell segmentation mostly used in the previous researches. Two-layer classification model is developed in this system, block pattern recognition and specimen pattern recognition; the staining pattern of the whole well is determined on the basis of the classification of its blocks through classifier fusion rules, such as majority rule, weighted majority rule and weighted sum rule. Relying on the results of block pattern classification, experiments on the whole images show that classifier fusion rules are able to identify the staining patterns of the whole well with a total accuracy of about 94.62%. Moreover, to solve the problem resulting from uneven distribution of blocks in the original classification system, an advanced multi-level classification with two stratification criteria, uniform stratification criterion and density stratification criterion, which divide blocks from block segmentation into several groups, is proposed to classify the staining patterns of HEp-2 cell images, which efficiently improves the classification performance.