Binary Grey Wolf Optimizer based Feature Selection for Nucleolar and Centromere Staining Pattern Classification in Indirect Immunofluorescence Images.

In this work, an attempt is made to distinguish nucleolar and centromere staining patterns using Bag-of-Keypoint Features (BoKF) model and Binary Grey Wolf Optimization (BGWO) based feature selection. Fluorescent staining patterns are produced by Indirect Immunofluorescence (IIF) Imaging and the patterns considered for this study are taken from a publicly available online database. The IIF images are pre-processed using edge-aware local contrast enhancement method. The contrast enhanced images are subjected to BoKF framework and Speeded up Robust Feature keypoints are extracted. Further, the most significant features are identified using BGWO and are fed to k-Nearest Neighbor (kNN) for classification. The results show that the BGWO features are able to classify the nucleolar and centromere patterns with an average accuracy of 91.6%. Results also indicate that the prominent features obtained using BGWO can improve the discrimination performance of IIF staining patterns. Hence it appears that the BGWO based feature selection could enable the computer aided diagnosis of autoimmune diseases.