Residual grounding transformer network for terrain recognition on the lunar surface.

It is of paramount importance for a rover running on an extraterrestrial body surface to recognize the dangerous zones autonomously. This automation is inevitable due to the communication delay. However, as far as we know, there are few annotated terrain recognition datasets for extraterrestrial bodies. Furthermore, the lack of datasets hinders the training and evaluation of recognition algorithms. Therefore, we first built the Chang'e 3 Terrain Recognition (CE3TR) Dataset to address terrain recognition and semantic segmentation problems on the lunar surface. The moon is one of the nearest celestial bodies to the earth; our work is geared towards extraterrestrial bodies. The images of our dataset are captured by the Yutu moon rover, which can retain the real illumination condition and terrain environment on the moon. A Residual Grounding Transformer Network (RGTNet) is also proposed to find out unsafe areas like rocks and craters. The residual grounding transformer is introduced to facilitate cross-scale interactions of different level features. A local binary pattern feature fusion module is another notable part of the RGTNet, which contributes to extracting the boundaries of different obstacles. We also present the ability of new loss, called smooth intersection over union loss, to mitigate overfitting. To evaluate RGTNet, we have conducted extensive experiments on our CE3TR Dataset. The experimental results demonstrate that our model can recognize risky terrain readily and outperforms other state-of-the-art methods.

Semi-supervised point consistency network for retinal artery/vein classification.

Recent deep learning methods with convolutional neural networks (CNNs) have boosted advance prosperity of medical image analysis and expedited the automatic retinal artery/vein (A/V) classification. However, it is challenging for these CNN-based approaches in two aspects: (1) specific tubular structures and subtle variations in appearance, contrast, and geometry, which tend to be ignored in CNNs with network layer increasing; (2) limited well-labeled data for supervised segmentation of retinal vessels, which may hinder the effectiveness of deep learning methods. To address these issues, we propose a novel semi-supervised point consistency network (SPC-Net) for retinal A/V classification. SPC-Net consists of an A/V classification (AVC) module and a multi-class point consistency (MPC) module. The AVC module adopts an encoder-decoder segmentation network to generate the prediction probability map of A/V for supervised learning. The MPC module introduces point set representations to adaptively generate point set classification maps of the arteriovenous skeleton, which enjoys its prediction flexibility and consistency (i.e. point consistency) to effectively alleviate arteriovenous confusion. In addition, we propose a consistency regularization between the predicted A/V classification probability maps and point set representations maps for unlabeled data to explore the inherent segmentation perturbation of the point consistency, reducing the need for annotated data. We validate our method on two typical public datasets (DRIVE, HRF) and a private dataset (TR280) with different resolutions. Extensive qualitative and quantitative experimental results demonstrate the effectiveness of our proposed method for supervised and semi-supervised learning.

18α-Glycyrrhetinic acid (GA) ameliorates fructose-induced nephropathy in mice by suppressing oxidative stress, dyslipidemia and inflammation.

Excessive fructose (FRU) intake can result in insulin resistance and metabolic disorder, which are related to renal injury.18α-Glycyrrhetinic acid (GA) is a bioactive component mainly extracted from Glycyrrhiza radix, and has anti-oxidant and anti-inflammatory activities. However, its effects on FRU-induced renal injury still remain unclear. In this study, we found that 18α-GA treatments could significantly ameliorate the cell viability in FRU-treated tubule epithelial cells, accompanied with improved mitochondrial membrane potential. Furthermore, reactive oxygen species (ROS) accumulation in FRU-stimulated cells was markedly reduced by 18α-GA, which were associated with the activation of nuclear factor (erythroid-derived-2)-like 2 (Nrf-2) and the blockage of MAPKs signaling. Additionally, dyslipidemia detected in FRU-treated cells was greatly inhibited by 18α-GA. We also found that 18α-GA significantly ameliorated FRU-induced inflammation in cells through reducing the expression of pro-inflammatory cytokines and chemokine. The anti-inflammatory effects regulated by 18α-GA were mainly related to the repression of nuclear factor-κB(NF-κB) signaling. Furthermore, the protective effects of 18α-GA against ROS production, lipid accumulation and inflammation were verified in renal tissues from FRU-challenged mice, consequently improving metabolic disorder and kidney injury. Taken together, these findings demonstrated that 18α-GA exerted renal protective effects through reducing oxidative stress, lipid deposition and inflammatory response, and thus could be considered as a promising therapeutic strategy for metabolic stress-induced kidney injury.

Sequential vessel segmentation via deep channel attention network.

Accurately segmenting contrast-filled vessels from X-ray coronary angiography (XCA) image sequence is an essential step for the diagnosis and therapy of coronary artery disease. However, developing automatic vessel segmentation is particularly challenging due to the overlapping structures, low contrast and the presence of complex and dynamic background artifacts in XCA images. This paper develops a novel encoder-decoder deep network architecture which exploits the several contextual frames of 2D+t sequential images in a sliding window centered at current frame to segment 2D vessel masks from the current frame. The architecture is equipped with temporal-spatial feature extraction in encoder stage, feature fusion in skip connection layers and channel attention mechanism in decoder stage. In the encoder stage, a series of 3D convolutional layers are employed to hierarchically extract temporal-spatial features. Skip connection layers subsequently fuse the temporal-spatial feature maps and deliver them to the corresponding decoder stages. To efficiently discriminate vessel features from the complex and noisy backgrounds in the XCA images, the decoder stage effectively utilizes channel attention blocks to refine the intermediate feature maps from skip connection layers for subsequently decoding the refined features in 2D ways to produce the segmented vessel masks. Furthermore, Dice loss function is implemented to train the proposed deep network in order to tackle the class imbalance problem in the XCA data due to the wide distribution of complex background artifacts. Extensive experiments by comparing our method with other state-of-the-art algorithms demonstrate the proposed method's superior performance over other methods in terms of the quantitative metrics and visual validation. To facilitate the reproductive research in XCA community, we publicly release our dataset and source codes at https://github.com/Binjie-Qin/SVS-net.