Combining convolutional neural networks and self-attention for fundus diseases identification.

Early detection of lesions is of great significance for treating fundus diseases. Fundus photography is an effective and convenient screening technique by which common fundus diseases can be detected. In this study, we use color fundus images to distinguish among multiple fundus diseases. Existing research on fundus disease classification has achieved some success through deep learning techniques, but there is still much room for improvement in model evaluation metrics using only deep convolutional neural network (CNN) architectures with limited global modeling ability; the simultaneous diagnosis of multiple fundus diseases still faces great challenges. Therefore, given that the self-attention (SA) model with a global receptive field may have robust global-level feature modeling ability, we propose a multistage fundus image classification model MBSaNet which combines CNN and SA mechanism. The convolution block extracts the local information of the fundus image, and the SA module further captures the complex relationships between different spatial positions, thereby directly detecting one or more fundus diseases in retinal fundus image. In the initial stage of feature extraction, we propose a multiscale feature fusion stem, which uses convolutional kernels of different scales to extract low-level features of the input image and fuse them to improve recognition accuracy. The training and testing were performed based on the ODIR-5k dataset. The experimental results show that MBSaNet achieves state-of-the-art performance with fewer parameters. The wide range of diseases and different fundus image collection conditions confirmed the applicability of MBSaNet.

CAN Canopy Addition of Nitrogen Better Illustrate the Effect of Atmospheric Nitrogen Deposition on Forest Ecosystem?

Increasing atmospheric nitrogen (N) deposition could profoundly impact community structure and ecosystem functions in forests. However, conventional experiments with understory addition of N (UAN) largely neglect canopy-associated biota and processes and therefore may not realistically simulate atmospheric N deposition to generate reliable impacts on forest ecosystems. Here we, for the first time, designed a novel experiment with canopy addition of N (CAN) vs. UAN and reviewed the merits and pitfalls of the two approaches. The following hypotheses will be tested: i) UAN overestimates the N addition effects on understory and soil processes but underestimates those on canopy-associated biota and processes, ii) with low-level N addition, CAN favors canopy tree species and canopy-dwelling biota and promotes the detritus food web, and iii) with high-level N addition, CAN suppresses canopy tree species and other biota and favors rhizosphere food web. As a long-term comprehensive program, this experiment will provide opportunities for multidisciplinary collaborations, including biogeochemistry, microbiology, zoology, and plant science to examine forest ecosystem responses to atmospheric N deposition.

Methane uptake in forest soils along an urban-to-rural gradient in Pearl River Delta, South China.

We investigated soil CH4 fluxes from six forests along an urban-to-rural gradient in Guangzhou City metropolitan area, South China. The most significant CH4 consumption was found in the rural site, followed by suburban, and then urban forest sites. The rates of CH4 uptake were significantly higher (by 38% and 44%, respectively for mixed forest and broadleaf forest) in the rural than in the urban forest site. The results indicate that soil water filled pore space (WFPS) is the primary factor for controlling CH4 consumption in subtropical forests. The reductions of soil CH4 uptake in urban forests were also influenced by the higher rates of atmospheric nitrogen (N) deposition and increases in soil nitrate (NO3(-)) and aluminum (Al(3+)) contents as a result of urbanization. Results from this work suggest that environmental changes associated with urbanization could decrease soil CH4 consumption in subtropical forests and potentially contribute to increase of atmospheric CH4 concentration.

Responses of CO(2), N(2)O and CH(4) fluxes between atmosphere and forest soil to changes in multiple environmental conditions.

To investigate the effects of multiple environmental conditions on greenhouse gas (CO2 , N2 O, CH4 ) fluxes, we transferred three soil monoliths from Masson pine forest (PF) or coniferous and broadleaved mixed forest (MF) at Jigongshan to corresponding forest type at Dinghushan. Greenhouse gas fluxes at the in situ (Jigongshan), transported and ambient (Dinghushan) soil monoliths were measured using static chambers. When the transported soil monoliths experienced the external environmental factors (temperature, precipitation and nitrogen deposition) at Dinghushan, its annual soil CO2 emissions were 54% in PF and 60% in MF higher than those from the respective in situ treatment. Annual soil N2 O emissions were 45% in PF and 44% in MF higher than those from the respective in situ treatment. There were no significant differences in annual soil CO2 or N2 O emissions between the transported and ambient treatments. However, annual CH4 uptake by the transported soil monoliths in PF or MF was not significantly different from that at the respective in situ treatment, and was significantly lower than that at the respective ambient treatment. Therefore, external environmental factors were the major drivers of soil CO2 and N2 O emissions, while soil was the dominant controller of soil CH4 uptake. We further tested the results by developing simple empirical models using the observed fluxes of CO2 and N2 O from the in situ treatment and found that the empirical models can explain about 90% for CO2 and 40% for N2 O of the observed variations at the transported treatment. Results from this study suggest that the different responses of soil CO2 , N2 O, CH4 fluxes to changes in multiple environmental conditions need to be considered in global change study.