Automatic diagnosis of autoimmune diseases by classifying HEp-2 cells

• Published in: 2017 13th International Computer Engineering Conference (ICENCO) pp. 208 - 212
• Main Authors: Kasem, Maryam A. G., Hadhoud, Marwa M. A., ELdesoky, Mohammed A. A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-12-2017
• Subjects: autoimmune diseases; Discrete cosine transforms; Diseases; Feature extraction; Fluorescence; HEP-2 Cells; Immune system; Shape; Support vector machines; Texture Features
• ISSN: 2475-2320
• DOI: 10.1109/ICENCO.2017.8289789

Immune system failure can cause many human autoimmune diseases which make the body produce antibodies in the blood, which attack human antigens not foreign antigens and it can cause many serious and chronic illnesses. This phenomenon results in an auto-immune response which generates the auto-antibodies (ANAs) during this response. This auto-antibodies (ANAs) may be identified through the indirect immunofluorescence test (IIF). But unfortunately, today the IIF is still a subjective method as well as time and labor intensive. Thus, to address these problems, computer Aided Diagnostic (CAD) systems have been developed which automatically classify a human epithelial cell type-2 (HEp-2) cell images into one of its known patterns. This paper proposes a system based on two phases. The first phase is feature extraction using morphological features, texture features, Discrete Cosine Transform (DCT) coefficients and combinations of these three categories of features. To identify the HEP-2 cell images to its different classes. The second phase is classification by using K-Nearest Neighbor (KNN) and Forward Neural Network (FNN) classifier. The applied experiments achieved average accuracy about 94.72%.