Contrastive self-supervised learning for diabetic retinopathy early detection.

Diabetic Retinopathy (DR) is the major cause of blindness, which seriously threatens the world's vision health. Limited medical resources make early diagnosis and a large-scale screening of DR difficult. Most of the current automatic diagnostic methods are mostly based on deep learning and large-scale labeled data. However, the insufficiency of manual annotations for medical images still is a great challenge of training deep neural networks. Self-supervised learning methods are proposed to learn general features from dataset without manual annotations. Inspired by this, we proposed a deep learning based DR classification model (SimCLR-DR). In this paper, we first use contrastive self-learning algorithm to pre-train the encoder based on convolution network with unlabeled retinal images, then retrain the encoder with classifier on a small annotated training data to detect referable DR. The experimental results on Kaggle dataset show that this proposed method can overcome the training data insufficiency problem and performs better than transfer learning. SimCLR-DR is a good beginning for other deep learning based medical image detection approaches facing the challenge of insufficient annotated data. Figure presents an overview of the proposed framework, which contains three main steps: (i) Data preprocessing; (ii) Pretext task of SimCLR-DR based on contrastive learning; (iii) Downstream Task of SimCLRDR based on CNN.

Integrated effects of polymer type, size and shape on the sinking dynamics of biofouled microplastics.

Sinking of microplastics (MPs) after biofouling is considered an important mechanisms responsible for the downward transport/sedimentation of MPs in the ocean and freshwaters. Previous studies demonstrated MP sinking caused by an increase in the composite density of MPs after biofouling, while MPs with smaller size or shapes with higher surface area to volume ratios (SA:V), such as films, are speculated to sink faster. In this study, we designed an in situ microcosm to simulate the ambient environmental conditions experienced by floating MPs to elucidate the biofouling and sinking of polyethylene (PE), polypropylene (PP), and expanded-polystyrene (EPS) MPs of various sizes and shapes. Our results showed smaller PE and PP MP granules sank faster than large ones. Even EPS granules of 100 µm diameter, having a much lower density (0.02 mg/mm3) than water, started to sink after 2 weeks of biofouling. Moreover, PE film and fiber MPs with higher SA:V did not sink faster than PE MP granules of the same mass, implying that mechanisms other than SA:V, such as fouling contact area and drag coefficient, play a role in the regulation of biofouling and sinking of MPs.

Estuarine salinity gradient governs sedimentary bacterial community but not antibiotic resistance gene profile.

Antibiotic resistance gene (ARG) pollution in estuarine environment has drawn great attention, and it is not clear if the physical and chemical parameters such as salinity, total organic carbon, total nitrogen, total phosphorus and antibiotics affects the distribution of ARGs. Herein, we deciphered the ARG profiles and microbial community compositions in sediments from Jiulong River Estuary (JRE) and Min River Estuary (MRE) of China using high-throughput sequencing-based metagenomics analysis. Furthermore, we explored the influence of salinity on bacterial community and ARG profiles. The results showed that Proteobacteria, Chloroflexi, and Acidobacteria were the dominant phyla in these two estuaries. The abundance of ARGs ranged from 1.05 × 10-1 to 2.93 × 10-1 copy of ARG per copy of 16S rRNA gene in all the sediment samples and the profiles of ARGs presented similar patterns in two estuaries. Multidrug resistance genes were the dominant ARG types in both estuaries, with an overall abundance of 2.39 × 10-2-1.07 × 10-1 copy of ARG per copy of 16S rRNA gene, followed by genes conferring resistance to bacitracin and macrolide-lincosamide-streptogramin. Salinity was an important influencing factor on the bacterial community but not on the ARG profiles. Instead, stochastic processes exerted the main influence on the distribution of ARGs. The comparison of ARG profiles among estuary sediments, marine sediments, and samples from anthropogenic pollution environments revealed remarkable similarity of ARG profiles between samples from estuary sediments and those from municipal wastewater treatment plants. These results suggested that the complex emission of anthropogenic pollution could cause the stochastic ecological pattern of ARGs.

Microplastics in seawater and two sides of the Taiwan Strait: Reflection of the social-economic development.

Microplastic abundance, distribution and source characteristics were investigated for the surface seawaters from the Taiwan Strait as well as those of sediments along its west and east coasts. The microplastic abundances were in the range of 28-208 (mean 90) and 10-246 (mean 69) items/kg (d.w.) along the west and east coasts respectively. The higher microplastic abundance on the west coast might be related to the different local economic development, population, land-use and other human activities. Combined with microplastic pollution and socio-economic development, regression analysis results showed that urbanization level is negatively correlated with foams while positively correlated with fibers. This study, as the first report of microplastics in the Taiwan Strait, suggested further research on microplastics cross-strait transportation and the relationship with economic developments.

Enhanced Sinks of Polycyclic Aromatic Hydrocarbons Due to Kuroshio Intrusion: Implications on Biogeochemical Processes in the Ocean-Dominated Marginal Seas.

The biogeochemical processes of polycyclic aromatic hydrocarbons (PAHs) in the South China Sea (SCS) are influenced by the exchanges of water masses, energies, and materials between this marginal sea and the Pacific Ocean. To investigate the impact of oceanic water intrusion on semivolatile compounds, we collected seawater samples in the Western Pacific, northern, and central SCS in 2017 and analyzed for dissolved PAHs. PAH concentrations in the water columns of the Pacific Ocean and SCS were 1.7-11 and 1.1-7.3 ng L-1, respectively, showing spatial distinctions in terms of the composition and source characteristics. A common depletion for three-ring PAHs was found in the northern SCS by comparing the modeling results of conservative mixing by Kuroshio intrusion. Kuroshio water increased the levels of temperature, dissolved oxygen, and nutrients when intruding into the northern SCS and was likely to enhance the bioavailability of PAHs and stimulate their biodegradation process. In the water column, the most effective layer under the Kuroshio intrusion impact is different for three- and four-ring PAHs, where the three-ring PAHs' depletion was most significant at the surface; however, for four-ring PAHs, that was at the deep chlorophyll maximum layer. This study highlighted the effect of ocean currents on PAHs for their water-column processes both from physical and biogeochemical perspectives.

Dynamics of size-fractionated bacterial communities during the coastal dispersal of treated municipal effluents.

Everyday huge amount of treated municipal wastewater is discharged into the coastal seawater. However, microbial biomarkers for the municipal effluent instead of the fecal species from raw sewage have not been proposed. Meanwhile, bacterial taxa for degrading large amounts of input organics have not been fully understood. In this study, raw effluent and serial water samples were collected from the coastal dispersal of two sewage treatment plants in Xiamen, China. Free-living (FL) and particle-associated (PA) bacterial communities were analyzed via high-throughput sequencing of 16S rRNA gene and quantitative PCR to measure bacterial abundance. The PA bacterial communities in our samples exhibited higher cell abundance, alpha diversity, and population dynamics than the FL bacterial communities, which supports greater environmental significance of the PA bacterial communities. Two non-fecal but typical genera in activated sludge, Zoogloea and Dechloromonas, exhibited decreased but readily detectable abundance along the effluent dispersal distance. Furthermore, the dominating microbial species near the outfalls were related to well-known marine indigenous taxa, such as SAR11 clade, OM60 clade, low-GC Actinobacteria, and unclassified Flavobacteriales, as well as the less understood taxa like Pseudohongiella and Microbacteriaceae. It is interesting that these taxa exhibited two types of correlation patterns with COD concentration. Our study suggested Zoogloea as a potential indicator of municipal effluents and also proposed potential utilizers of residual effluent COD in marine environments.