data augmentation by generative adversarial networks for white blood cell image classification

Deep learning based algorithms have shown agreat success on different tasks including image classification.one of the requirements of implementing deep learningapproaches is availability of large-scale datasets. however, thelack of big medical datasets due to the difficulties in recordingthese kinds of data, is one of the major problems inimplementing deep learning approaches. therefore, dataaugmentation has become an important step for increasing thenumber of data samples. image rotating in different angles,horizontal and vertical flipping is one of the popular imagedata augmentation methods. however, the generated imagesare so similar to the original ones. recently, generativeadversarial neural networks (gans) have been proposed aspowerful methods for generating new data samples. in thisarticle, we explore image augmentation by gan structures tobe used in leukemia diagnosis task. to this end, a deepconvolutional gan is considered for generating white bloodcell images to increase the number of image samples ofallidb database. then, a deep convolutional neuralnetwork is applied on the augmented dataset to classify theimages as normal or leukemia. experimental results verify thatby implementing gan approach for image augmentation wecan achieve to 84%, classification accuracy which is 10%improvement with respect to the common augmentationmethod.