RESUMO
The detection of the optic disc (OD) and fovea is essential to many automatic diagnosis systems for retinal diseases. The single shot multibox detector (SSD) can generate predictions from feature maps of various resolutions, which has not been introduced into the OD and fovea detection. To enhance the detection performance, we propose an improved SSD network, which has strengthened information flow enabled by the dense connections. The proposed method can achieve multiscale detection of the OD and fovea with strengthened feature propagation. Extensive experiments on the publicly available Messidor database and local fundus images are performed to evaluate the performance of the proposed method. Compared with seven types of representative solutions in the Messidor database, the proposed method can achieve competitive performance compared to state-of-the-art algorithms. Furthermore, the proposed method is applied to the excyclotropia screening. The screening results demonstrate promising application prospects for the proposed method in medical practice.
Assuntos
Exotropia/diagnóstico por imagem , Fóvea Central/diagnóstico por imagem , Interpretação de Imagem Assistida por Computador/instrumentação , Redes Neurais de Computação , Disco Óptico/diagnóstico por imagem , Reconhecimento Automatizado de Padrão/métodos , Algoritmos , Bases de Dados Factuais , Humanos , Reprodutibilidade dos Testes , Sensibilidade e EspecificidadeRESUMO
The segmentation of blood vessels in retinal images is crucial to the diagnosis of many diseases. We propose a deep learning method for vessel segmentation based on an encoder-decoder network combined with squeeze-and-excitation connection and atrous spatial pyramid pooling. In our implementation, the atrous spatial pyramid pooling allows the network to capture features at multiple scales, and the high-level semantic information is combined with low-level features through the encoder-decoder architecture to generate segmentations. Meanwhile, the squeeze-and-excitation connections in the proposed network can adaptively recalibrate features according to the relationship between different channels of features. The proposed network can achieve precise segmentation of retinal vessels without hand-crafted features or specific post-processing. The performance of our model is evaluated in terms of visual effects and quantitative evaluation metrics on two publicly available datasets of retinal images, the Digital Retinal Images for Vessel Extraction and Structured Analysis of the Retina datasets, with comparison to 12 representative methods. Furthermore, the proposed network is applied to vessel segmentation on local retinal images, which demonstrates promising application prospect in medical practices.
Assuntos
Algoritmos , Processamento de Imagem Assistida por Computador , Vasos Retinianos/diagnóstico por imagem , Humanos , Modelos Teóricos , Curva ROCRESUMO
Submicron droplets are ubiquitous in nature and widely applied in fields such as biomedical diagnosis and therapy, oil recovery and energy conversion, among others. The submicron droplets are kinetically stable, their submicron size endows them with good mobility in highly constricted pathways, and the high surface-to-volume ratio allows effective loading of chemical components at the interface and good heat transfer performance. Conventional generation technology of submicron droplets in bulk involves high energy input, or relies on chemical energy released from the system. Microfluidic methods are widely used to generate highly monodispersed micron-sized or bigger droplets, while downsizing to the order of 100 nm was thought to be challenging because of sophisticated nanofabrication. In this review, we summarize the microfluidic methods that are promising for the generation of submicron droplets, with an emphasize on the device fabrication, operational condition, and resultant droplet size. Microfluidics offer a relatively energy-efficient and versatile tool for the generation of highly monodisperse submicron droplets.
RESUMO
Micron- and submicron-sized droplets have extensive applications in biomedical diagnosis and drug delivery. Moreover, accurate high-throughput analysis requires a uniform droplet size distribution and high production rates. Although the previously reported microfluidic coflow step-emulsification method can be used to generate highly monodispersed droplets, the droplet diameter (d) is constrained by the microchannel height (b), dâ³3b, while the production rate is limited by the maximum capillary number of the step-emulsification regime, impeding emulsification of highly viscous liquids. In this paper, we report a novel, gas-assisted coflow step-emulsification method, where air serves as the innermost phase of a precursor hollow-core air/oil/water emulsion. Air gradually diffuses out, producing oil droplets. The size of the hollow-core droplets and the ultrathin oil layer thickness both follow the scaling laws of triphasic step-emulsification. The minimal droplet size attains d≈1.7b, inaccessible in standard all-liquid biphasic step-emulsification. The production rate per single channel is an order-of-magnitude higher than that in the standard all-liquid biphasic step-emulsification and is also superior to alternative emulsification methods. Due to low gas viscosity, the method can also be used to generate micron- and submicron-sized droplets of high-viscosity fluids, while the inert nature of the auxiliary gas offers high versatility.
RESUMO
The susceptibility to trace metals and legacy POPs is different between terrestrial and marine mammals. In this study, we established the first cell line from Indo-Pacific finless porpoises and compared the cellular responses of skin fibroblast cells from Pygmy killer whales, Pantropic spotted dolphins, Indo-Pacific finless porpoises, mice, and humans following exposure to copper, methylmercury, cadmium, PCB126, PCB153, and BDE47 to better understand the interspecies sensitivities of mammals to chemical pollutants. We conducted a risk assessment by comparing no-observed effect concentrations (NOEC), lowest-observed effect concentrations (LOEC), and half maximal effective concentrations (EC50) from cell viability assays and previously reported pollutant body burdens in mammals. Based on the in vitro data, Indo-Pacific finless porpoises were more sensitive to copper and methylmercury than other mammals. PCB153 exposure reduced cell viability in all mammals except humans, while PCB126 was more potent, with 13.33 µg/mL exposure reducing cell viability in all mammals. In contrast, BDE47 exposure reduced cell viability only in terrestrial mammals in addition to pantropic spotted dolphin. Based on the in vitro data and the natural context of metal concentrations, both methylmercury and cadmium posed a higher risk to cetaceans than human, while copper posed a lower risk to cetaceans. All three legacy POPs (PCB126, PCB153, and BDE47) posed minor risk to cetaceans for short-term exposure. This study demonstrated that a species-specific in vitro model may provide more accurate information on the potential risk of pollutants to mammals. However, due to the bioamplification of POPs and their potential impact on the endocrine system and immune system of cetaceans, risk assessment with long-term exposure with more in vitro models should be further studied.
Assuntos
Golfinhos , Poluentes Ambientais , Compostos de Metilmercúrio , Toninhas , Oligoelementos , Poluentes Químicos da Água , Humanos , Animais , Camundongos , Poluentes Químicos da Água/análise , Compostos de Metilmercúrio/metabolismo , Cobre/toxicidade , Cobre/metabolismo , Cádmio/metabolismo , Toninhas/metabolismo , Golfinhos/metabolismo , Oligoelementos/toxicidade , Oligoelementos/metabolismo , Poluentes Ambientais/metabolismo , FibroblastosRESUMO
The gut microbiome is a unique marker for cetaceans' health status, and the microbiome composition of their skin wounds can indicate a potential infection from their habitat. Our study provides the first comparative analysis of the microbial communities from gut regions and skin wounds of an individual Indo-Pacific finless porpoise (Neophocaena phocaenoides). Microbial richness increased from the foregut to the hindgut with variation in the composition of microbes. Fusobacteria (67.51% ± 5.10%), Firmicutes (22.00% ± 2.60%), and Proteobacteria (10.47% ± 5.49%) were the dominant phyla in the gastrointestinal tract, while Proteobacteria (76.11% ± 0.54%), Firmicutes (22.00% ± 2.60%), and Bacteroidetes (10.13% ± 0.49%) were the dominant phyla in the skin wounds. The genera Photobacterium, Actinobacillus, Vibrio, Erysipelothrix, Tenacibaculum, and Psychrobacter, considered potential pathogens for mammals, were identified in the gut and skin wounds of the stranded Indo-Pacific finless porpoise. A comparison of the gut microbiome in the Indo-Pacific finless porpoise and other cetaceans revealed a possible species-specific gut microbiome in the Indo-Pacific finless porpoise. There was a significant difference between the skin wound microbiomes in terrestrial and marine mammals, probably due to habitat-specific differences. Our results show potential species specificity in the microbiome structure and a potential threat posed by environmental pathogens to cetaceans.