MITOTIC HEp-2 CELL RECOGNITION USING LOCAL BINARY PATTERN (LBP) AND k-NEAREST NEIGHBOUR (k-NN) CLASSIFIER

Immune system produces autoantibody either randomly or as a result of an unknown material in the body itself. A failure of the body’s own defences against diseases has result to the auto immune diseases. Auto immune diseases will start attacking its own cell as they unable to differentiate between the foreign material and its own cell. An antinuclear antibody (ANA) is required as an indicator of the autoimmune process. IIF is the recommended technique for ANA detection in patient serum. IIF slides are observed by specialists reporting for the fluorescence intensity, staining pattern and looking for the mitotic cells. Indeed, the presence of the mitotic cells significantly important due to several key factors, first to guarantee the correctness of the slide preparation process and reported the staining pattern. Therefore the ability to detect mitotic cells is acquired to develop Computer-Aided Diagnosis (CAD) system in IIF to support the specialists form image acquisition up to image classification. This work aims to highlight the features of mitotic cells and to develop recognition algorithm for mitotic cell by incorporated Local Binary Pattern (LBP) and k-Nearest Neighbour to classified unlabelled image. A completed modelling of the LBP operator is proposed which is represented by its centre pixel and a local sign-magnitude transform (LDSMT). k-NN classifies unlabelled images based on the utmost majority samples in the feature space. This work involves five stages; image acquisition, pre-processing, feature extraction, distance measurement and classification.