资源环境科技发展态势分析平台

Screening mammography aims to identify breast cancer at earlier stages of the disease, when treatment can be more successful(1). Despite the existence of screening programmes worldwide, the interpretation of mammograms is affected by high rates of false positives and false negatives(2). Here we present an artificial intelligence (AI) system that is capable of surpassing human experts in breast cancer prediction. To assess its performance in the clinical setting, we curated a large representative dataset from the UK and a large enriched dataset from the USA. We show an absolute reduction of 5.7% and 1.2% (USA and UK) in false positives and 9.4% and 2.7% in false negatives. We provide evidence of the ability of the system to generalize from the UK to the USA. In an independent study of six radiologists, the AI system outperformed all of the human readers: the area under the receiver operating characteristic curve (AUC-ROC) for the AI system was greater than the AUC-ROC for the average radiologist by an absolute margin of 11.5%. We ran a simulation in which the AI system participated in the double-reading process that is used in the UK, and found that the AI system maintained non-inferior performance and reduced the workload of the second reader by 88%. This robust assessment of the AI system paves the way for clinical trials to improve the accuracy and efficiency of breast cancer screening.

Clinical trials using adult stem cells to regenerate damaged heart tissue continue to this day(1,2), despite ongoing questions of efficacy and a lack of mechanistic understanding of the underlying biological effect(3). The rationale for these cell therapy trials is derived from animal studies that show a modest but reproducible improvement in cardiac function in models of cardiac ischaemic injury(4,5). Here we examine the mechanistic basis for cell therapy in mice after ischaemia-reperfusion injury, and find that-although heart function is enhanced-it is not associated with the production of new cardiomyocytes. Cell therapy improved heart function through an acute sterile immune response characterized by the temporal and regional induction of CCR2(+) and CX3CR1(+) macrophages. Intracardiac injection of two distinct types of adult stem cells, cells killed by freezing and thawing or a chemical inducer of the innate immune response all induced a similar regional accumulation of CCR2(+) and CX3CR1(+) macrophages, and provided functional rejuvenation to the heart after ischaemia-reperfusion injury. This selective macrophage response altered the activity of cardiac fibroblasts, reduced the extracellular matrix content in the border zone and enhanced the mechanical properties of the injured area. The functional benefit of cardiac cell therapy is thus due to an acute inflammatory-based wound-healing response that rejuvenates the infarcted area of the heart.

Animals sense the environment through pathways that link sensory organs to the brain. In the visual system, these feedforward pathways define the classical feedforward receptive field (ffRF), the area in space in which visual stimuli excite a neuron(1). The visual system also uses visual context-the visual scene surrounding a stimulus-to predict the content of the stimulus(2), and accordingly, neurons have been identified that are excited by stimuli outside their ffRF(3-8). However, the mechanisms that generate excitation to stimuli outside the ffRF are unclear. Here we show that feedback projections onto excitatory neurons in the mouse primary visual cortex generate a second receptive field that is driven by stimuli outside the ffRF. The stimulation of this feedback receptive field (fbRF) elicits responses that are slower and are delayed in comparison with those resulting from the stimulation of the ffRF. These responses are preferentially reduced by anaesthesia and by silencing higher visual areas. Feedback inputs from higher visual areas have scattered receptive fields relative to their putative targets in the primary visual cortex, which enables the generation of the fbRF. Neurons with fbRFs are located in cortical layers that receive strong feedback projections and are absent in the main input layer, which is consistent with a laminar processing hierarchy. The observation that large, uniform stimuli-which cover both the fbRF and the ffRF-suppress these responses indicates that the fbRF and the ffRF are mutually antagonistic. Whereas somatostatin-expressing inhibitory neurons are driven by these large stimuli, inhibitory neurons that express parvalbumin and vasoactive intestinal peptide have mutually antagonistic fbRF and ffRF, similar to excitatory neurons. Feedback projections may therefore enable neurons to use context to estimate information that is missing from the ffRF and to report differences in stimulus features across visual space, regardless of whether excitation occurs inside or outside the ffRF. By complementing the ffRF, the fbRF that we identify here could contribute to predictive processing.

Feedback projections onto neurons of the mouse primary visual cortex generate a second excitatory receptive field that is driven by stimuli outside of the classical feedforward receptive field, with responses mediated by higher visual areas.

A video-based deep learning algorithm-EchoNet-Dynamic-accurately identifies subtle changes in ejection fraction and classifies heart failure with reduced ejection fraction using information from multiple cardiac cycles.

Accurate assessment of cardiac function is crucial for the diagnosis of cardiovascular disease(1), screening for cardiotoxicity(2) and decisions regarding the clinical management of patients with a critical illness(3). However, human assessment of cardiac function focuses on a limited sampling of cardiac cycles and has considerable inter-observer variability despite years of training(4,5). Here, to overcome this challenge, we present a video-based deep learning algorithm-EchoNet-Dynamic-that surpasses the performance of human experts in the critical tasks of segmenting the left ventricle, estimating ejection fraction and assessing cardiomyopathy. Trained on echocardiogram videos, our model accurately segments the left ventricle with a Dice similarity coefficient of 0.92, predicts ejection fraction with a mean absolute error of 4.1% and reliably classifies heart failure with reduced ejection fraction (area under the curve of 0.97). In an external dataset from another healthcare system, EchoNet-Dynamic predicts the ejection fraction with a mean absolute error of 6.0% and classifies heart failure with reduced ejection fraction with an area under the curve of 0.96. Prospective evaluation with repeated human measurements confirms that the model has variance that is comparable to or less than that of human experts. By leveraging information across multiple cardiac cycles, our model can rapidly identify subtle changes in ejection fraction, is more reproducible than human evaluation and lays the foundation for precise diagnosis of cardiovascular disease in real time. As a resource to promote further innovation, we also make publicly available a large dataset of 10,030 annotated echocardiogram videos.

Elucidating elementary mechanisms that underlie bacterial diversity is central to ecology(1,2) and microbiome research(3). Bacteria are known to coexist by metabolic specialization(4), cooperation(5) and cyclic warfare(6-8). Many species are also motile(9), which is studied in terms of mechanism(10,11), benefit(12,13), strategy(14,15), evolution(16,17) and ecology(18,19). Indeed, bacteria often compete for nutrient patches that become available periodically or by random disturbances(2,20,21). However, the role of bacterial motility in coexistence remains unexplored experimentally. Here we show that-for mixed bacterial populations that colonize nutrient patches-either population outcompetes the other when low in relative abundance. This inversion of the competitive hierarchy is caused by active segregation and spatial exclusion within the patch: a small fast-moving population can outcompete a large fast-growing population by impeding its migration into the patch, while a small fast-growing population can outcompete a large fast-moving population by expelling it from the initial contact area. The resulting spatial segregation is lost for weak growth-migration trade-offs and a lack of virgin space, but is robust to population ratio, density and chemotactic ability, and is observed in both laboratory and wild strains. These findings show that motility differences and their trade-offs with growth are sufficient to promote diversity, and suggest previously undescribed roles for motility in niche formation and collective expulsion-containment strategies beyond individual search and survival.

In mixed bacterial populations that colonize nutrient patches, a growth-migration trade-off can lead to spatial exclusion that provides an advantage to populations that become rare, thereby stabilizing the community.