he application of deep neural networks to medical imaging is an evolving research field (1,2). An artificial neural network consists of a set of simple processing units, artificial neurons, connected in a network, organized in layers, and trained with a backpropagation algorithm (3). The resulting computational model is able to learn representations of data with a high level of abstraction (4). Deep neural networks have been shown to achieve excellent performance on many natural computer vision tasks, which is advantageous for medical specialties such as radiology and dermatology (4,5). Previous work has indicated that the performance of deep learning algorithms is comparable to or even exceeds the performance of radiologists in detecting consolidation on chest radiographs (6), segmenting cysts in polycystic kidney disease on CT scans (7), and detecting pulmonary nodules on CT scans (8). Artificial intelligence (AI)-led independent reporting of imaging remains a controversial topic; however, many radiologists would agree that deep learning technology could be a valuable tool in improving workflow and workforce efficiency (9-11). The increasing clinical demands on radiology departments worldwide have challenged current service delivery models, particularly in publicly funded health care systems. In some settings, it may not be feasible to report all acquired radiographs in a timely manner, leading to large backlogs of unreported studies (12,13). For example, the United Kingdom estimates that, at any time, 330 000 patients are waiting more than 30 days for their reports (14). Therefore, alternative models of care should be explored, particularly for chest radiographs, which account for 40% of all diagnostic images worldwide (15). Better mechanisms for triaging abnormal versus normal chest radiographs and prioritization of abnormal radiographs (eg, according to the "criticality" of the

No abstract

Network Functions Virtualization (NFV) aims to transform network functions into software images, executed on standard, high-volume hardware. This paper focuses on the case in which a massive number of (tiny) network function instances are executed simultaneously on the same server and presents our experience in the design of the components that move the traffic across those functions, based on the primitives offered by the Intel DPDK framework. This paper proposes different possible architectures, it characterizes the resulting implementations, and it evaluates their applicability under different constraints.

No abstract

No abstract