Enhanced balancing GAN: minority-class image generation.

Generative adversarial networks (GANs) are one of the most powerful generative models, but always require a large and balanced dataset to train. Traditional GANs are not applicable to generate minority-class images in a highly imbalanced dataset. Balancing GAN (BAGAN) is proposed to mitigate this problem, but it is unstable when images in different classes look similar, e.g., flowers and cells. In this work, we propose a supervised autoencoder with an intermediate embedding model to disperse the labeled latent vectors. With the enhanced autoencoder initialization, we also build an architecture of BAGAN with gradient penalty (BAGAN-GP). Our proposed model overcomes the unstable issue in original BAGAN and converges faster to high-quality generations. Our model achieves high performance on the imbalanced scale-down version of MNIST Fashion, CIFAR-10, and one small-scale medical image dataset. https://github.com/GH920/improved-bagan-gp.

Clinical, cytological and microbiological evaluation of bronchoalveolar lavage in children: A referral hospital-based study.

INTRODUCTION: Diffuse lung diseases (DLD) in children involve a group of heterogeneous, rare disorders. In spite of the low diagnostic yield in pediatric DLD, bronchoalveolar lavage (BAL) can be used to diagnose specific disorders. There are few studies about microbial and cellular profiles of BAL samples in these patients. This study was conducted to evaluate the clinical, cytological and microbiological evaluation of BAL in children with DLD. METHODS: The clinical, cytological and microbiological profiles of BAL samples of all patients with DLD who underwent the fiberoptic bronchoscopy (FOB) at Children's Medical Center, an Iranian referral pediatrics Hospital during a year were evaluated. RESULTS: In 18 patients (18.4%) of the 98 cases studied, 22 pathogens were obtained as etiologic agents. The mean total cells count of BAL was 23.9 × 104 ± 12.9 × 104/ml. The mean percentages of cellular components were macrophages (70.2%), neutrophils (16.3%), lymphocytes (11.8%) and eosinophils (1.4%), respectively. The type of lung disease was significantly associated with the mean percentage of lymphocytes (p = 0.005) and the percentage of neutrophils (p = 0.042). CONCLUSION: FOB and BAL evaluation in combination with clinical and radiographic imaging data may be helpful for identifying of presumptive diagnosis of DLD in children.