Comparison of generative adversarial networks architectures forbiomedical images synthesis

Abstract

The article analyzes and compares the architectures of generativeadversarialnetworks. These networks are based on convolu-tional neural networks that are widely used for classification problems. Convolutional networks require a lot of training data to achieve the desired accuracy. Generativeadversarialnetworks are used for the synthesis of biomedical images in this work. Biomedi-cal images are widely used in medicine, especially in oncology. For diagnosis in oncology biomedical images are divided into three classes: cytological, histological, and immunohistochemical. Initial samples of biomedical images are very small. Getting trainingimages is a challenging and expensive process. A cytological training datasetwas used for the experiments. The article considers the most common architectures of generative adversarialnetworks suchas Deep Convolutional GAN (DCGAN), Wasserstein GAN (WGAN),Wasserstein GAN with gradient penalty (WGAN-GP), Boundary-seeking GAN (BGAN), Boundary equilibrium GAN (BEGAN). A typical GAN network architecture consists of a generator and discriminator. The generator and discriminator are based on the CNN network architecture.The algorithm of deep learning for image synthesis with the help ofgenerativeadversarialnet-worksis analyzed in the work. During the experiments, the following problems were solved. To increase the initial number of train-ingdata to the datasetapplied a set of affine transformations: mapping, paralleltransfer, shift, scaling, etc. Each of the architectures was trainedfor a certain numberof iterations. The selected architectures were compared by the training timeand image quality based on FID(FreshetInception Distance)metric. The experiments were implemented in Python language.Pytorch was used as a machine learning framework. Based on the used softwarea prototype software module for the synthesis of cytological imageswas developed. Synthesis of cytological images was performed on the basis of DCGAN, WGAN, WGAN-GP, BGAN, BEGAN architectures. Goog-le's online environment called Collaboratory was used for the experimentsusing Nvidia Tesla K80 graphics processor

Methodology for image retrieval based on binary spacepartitioning and perceptual image hashing

“…Recently, the most popular is the search for visual features using deep neural networks [22,23], [24]. This method is used both to search for global features and to search for local ones [25].…”

“…In the field of image search, as a rule, two types of binary space partition trees are used: kd-tree and vp-tree. Based on the works [24,25], within the framework of this technique, it was decided to use a vp-tree as a binary space partition tree. It is assumed that the use of the vp-tree will achieve better image search time compared to the kd-tree.…”

Generating High-Resolution Chest X-ray Images Using CGAN

“…The objective introduces two more hyper-parameters to keep a balance during the Entropy of the training procedure and conduct a theoretical analysis for selecting appropriate hyper-parameters. Berezsky et al [14] present conditional Deep Convolution GAN-particle swarm optimization (cDCGAN-PSO), a version of algorithm of GANs to generate chest X-rays images. cDCGANs can synthesize multiple classes from X-rays images, the limitation is the number of generated images is limited which generates only 600 images for the COVID-19 class.…”

“…The experiments of the proposed model and other methodologies are presented in this section, Table 4 comparison between the proposed model and previously reviewed models depending on the FID metric. Berezsky et al [14] Shmelkov et al [15] shuang et al [16] Waheed et al [17] Loey et al [18] proposed model…”