In this work, we present a novel approach for training Generative Adversarial Networks (GANs). Using the attention maps produced by a Teacher-Network we are able to improve the quality of the generated images as well as perform weakly object localization on the generated images. To this end, we generate images of HEp-2 cells captured with Indirect Imunofluoresence (IIF) and study the ability of our network to perform a weakly localization of the cell. Firstly, we demonstrate that whilst GANs can learn the mapping between the input domain and the target distribution efficiently, the discriminator network is not able to detect the regions of interest. Secondly, we present a novel attention transfer mechanism which allows us to enforce the discriminator to put emphasis on the regions of interest via transfer learning. Thirdly, we show that this leads to more realistic images, as the discriminator learns to put emphasis on the area of interest. Fourthly, the proposed method allows one to generate both images as well as attention maps which can be useful for data annotation e.g in object detection.
Autoimmune diseases are proven to be connected with the occurrence of autoantibodies in patient serum. Antinuclear autoantibodies (ANAs) identification can be accomplished in a laboratory using indirect immunofluorescence (IIF) imaging. ANAs are characterized by specific "visual" patterns on a humane epithelial cell line (HEp-2). The identification stage is usually done by trained and highly qualified physicians through visual inspection of slides using a fluorescence microscope. The presence of subjectivity in the identification process, the interobserver variability, the increasing demand of highly specialized personnel, suggest that a realization of an automatic classification system is of great significance for the field of autoimmune diseases diagnosis. Moreover CAD systems can be used in a collaborative scheme in order to augment the physicians' capabilities. In this paper a system for automatic classification of staining patterns on single-cell fluorescence images is proposed. Our method utilizes morphological features extracted from a set of binary images derived via multi-level thresholding of fluorescence images. Furthermore, a modified version of Uniform Local Binary Patterns descriptor is incorporated in order to capture local textural information. The classification is performed using a non-linear SVM Classifier. The proposed method is evaluated using a publicly available dataset, recently released for the purposes of HEP-2 Cells classification competition at ICPR 2012, achieving up to 95.9% overall classification accuracy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.