Early detection and classification of pulmonary nodules using computer-aided diagnosis (CAD) systems is useful in reducing mortality rates of lung cancer. In this paper, we propose a new deep learning method to improve classification accuracy of pulmonary nodules in computed tomography (CT) scans. Our method uses a novel 15-layer 2D deep convolutional neural network architecture for automatic feature extraction and classification of pulmonary candidates as nodule or nonnodule. Focal loss function is then applied to the training process to boost classification accuracy of the model. We evaluated our method on the LIDC/IDRI dataset extracted by the LUNA16 challenge. The experiments showed that our deep learning method with focal loss is a high-quality classifier with an accuracy of 97.2%, sensitivity of 96.0%, and specificity of 97.3%.

International audienceVirtualized cloud infrastructures are very popular as they allow resource mutualization and therefore cost reduction. For cloud providers, minimizing the number of used resources is one of the main services that such environments must ensure. Cloud customers are also concerned with the minimization of used resources in the cloud since they want to reduce their invoice. Thus, resource management in the cloud should be considered by the cloud provider at the virtualization level and by the cloud customers at the application level. Many research works investigate resource management strategies in these two levels. Most of them study virtual machine consolidation (according to the virtualized infrastructure utilization rate) at the virtualized level and dynamic application sizing (according to its workload) at the application level. However, these strategies are studied separately. In this article, we show that virtual machine consolidation and dynamic application sizing are complementary. We show the efficiency of the combination of these two strategies, in reducing resource usage and keeping an application’s Quality of Service. Our demonstration is done by comparing the evaluation of three resource management strategies (implemented at the virtualization level only, at the application level only, or complementary at both levels) in a private cloud infrastructure, hosting typical JEE web applications (evaluated with the RUBiS benchmark)