Studying the Applicability of Generative Adversarial Networks on HEp-2 Cell Image Augmentation

The Anti-Nuclear Antibodies (ANAs) testing is the primary serological diagnosis screening test for autoimmune diseases. ANAs conducted mainly by Indirect Immunofluorescence (IIF) on Human Epithelial cell-substrate (HEp-2) protocol. However, due to its high variability, human-subjectivity, and low throughput, there an insistent need develop efficient Computer-Aided Diagnosis system (CADs) automate this Many recently proposed Convolutional Neural Networks (CNNs) demonstrated promising results in HEp-2 cell image classification, which main task of HE-p2 IIF lack large labeled datasets still challenge field. This work provides a detailed study applicability using generative adversarial networks (GANs) algorithms as augmentation method. Different types GANs were employed synthesize images address data scarcity problem. For systematic comparison, empirical quantitative metrics implemented evaluate different GAN models' performance learning real representations. showed that though visual similarity with images, GANs' capacity generate diverse limited. deficiency generated diversity found be crucial impact when used standalone method augmentation. combining limited-size GANs-generated classic improves classification accuracy across variants CNNs. Our competitive overall mean class task.