Search

Visual-based lane detection is a critical task for autonomous driving systems. Existing methods yield good performance in lane detection in most cases, but they often fail to provide satisfactory results in challenging scenarios such as heavy occlusions or extreme lighting conditions. Theoretically, learning lane features from global context information is considered to be the most effective method. Motivated by this observation, a novel attention-fused instance segmentation network named AFLaneNet is proposed to aggregate contextual information and pay attention to lane features for real-time lane detection. Firstly, an attention gate is incorporated to effectively fuse shallow and deep features to eliminate aliasing effects caused by feature fusion. Then, the disturbance information affecting the lane features is reduced by combining spatial and channel attention mechanism. Moreover, the Lane Self Attention (LSA) module is designed as an attention-guided auxiliary part to enhance the lane features by extracting global context information. Experimental results on two widely used datasets (CULane and TuSimple) demonstrate that our approach shows remarkable performance in terms of both efficacy and efficiency compared with those existing segmentation methods for lane detection, with good robustness under complex lighting and occlusion conditions. [ABSTRACT FROM AUTHOR]

Much work has been done to improve visual SLAM systems by integrating point, line, and plane features into the bundle adjustment model, however little attention has been paid to explore the associations between these spatial features that could be used to achieve better performance. This study proposes a multimodal feature association-based stereo SLAM method. Firstly, a method to extract point and line features from stereo images and estimate plane features through the line features is proposed. Then, the corresponding mathematic models for representing the association relations between different features are given. After that, the association relations are integrated into the back-end optimization model to improve system robustness and accuracy by adjusting reprojection errors with confidence weights. Finally, comparison tests are performed to evaluate the performance of the proposed method with the mainstream stereo SLAM systems using the EuRoC and KITTI datasets. The test results show that our approach not only generates semantic maps with higher fidelity but also provides a better positioning capability, with the positioning accuracy of the algorithm on the EuRoC and KITTI datasets improved by an average of 14.78% and 20.09%, respectively. [ABSTRACT FROM AUTHOR]

Since its first outbreak, Coronavirus Disease 2019 (COVID-19) has been rapidly spreading worldwide and caused a global pandemic. Rapid and early detection is essential to contain COVID-19. Here, we first developed a deep learning (DL) integrated radiomics model for end-to-end identification of COVID-19 using CT scans and then validated its clinical feasibility. We retrospectively collected CT images of 386 patients (129 with COVID-19 and 257 with other community-acquired pneumonia) from three medical centers to train and externally validate the developed models. A pre-trained DL algorithm was utilized to automatically segment infected lesions (ROIs) on CT images which were used for feature extraction. Five feature selection methods and four machine learning algorithms were utilized to develop radiomics models. Trained with features selected by L1 regularized logistic regression, classifier multi-layer perceptron (MLP) demonstrated the optimal performance with AUC of 0.922 (95% CI 0.856–0.988) and 0.959 (95% CI 0.910–1.000), the same sensitivity of 0.879, and specificity of 0.900 and 0.887 on internal and external testing datasets, which was equivalent to the senior radiologist in a reader study. Additionally, diagnostic time of DL-MLP was more efficient than radiologists (38 s vs 5.15 min). With an adequate performance for identifying COVID-19, DL-MLP may help in screening of suspected cases. [ABSTRACT FROM AUTHOR]

Aims and background: The COVID-19 outbreak spread in China and is a threat to the world. We reported on the epidemiological, clinical, laboratory, and radiological characteristics of children cases to help health workers better understand and provide timely diagnosis and treatment. Methods: Retrospectively, two research centers' case series of 67 consecutive hospitalized cases including 53 adult and 14 children cases with COVID-19 between 23 Jan 2020 and 15 Feb 2020 from Jinan and Rizhao were enrolled in this study. Epidemiological, clinical, laboratory, and radiological characteristics of children and adults were analyzed and compared. Results: Most cases in children were mild (21.4%) and conventional cases (78.6%), with mild clinical signs and symptoms, and all cases were of family clusters. Fever (35.7%) and dry cough (21.4%) were described as clinical manifestations in children cases. Dry cough and phlegm were not the most common symptoms in children compared with adults (p = 0.03). In the early stages of the disease, lymphocyte counts did not significantly decline but neutrophils count did in children compared with adults (p = 0.02). There was a lower level of CRP (p = 0.00) in children compared with adults. There were 8 (57.1%) asymptomatic cases and 6 (42.9%) symptomatic cases among the 14 children cases. The age of asymptomatic patients was younger than that of symptomatic patients (p = 0.03). Even among asymptomatic patients, 5 (62.5%) cases had lung injuries including 3 (60%) cases with bilateral involvement, which was not different compared with that of symptomatic cases (p = 0.58, p = 0.74). Conclusions: The clinical symptoms of children are mild, there is substantial lung injury even among children, but that there is less clinical disease, perhaps because of a less pronounced inflammatory response, and that the occurrence of this pattern appears to inversely correlate with age. [ABSTRACT FROM AUTHOR]

In this paper, we present a two-phase random-valued impulse noise removal algorithm based on local deviation index ( LDI) and edge-preserving regularization. In the first phase, we define an image statistic LDI. Then with image pixels' LDI values, the outlier candidates are identified. In the second phase, the image is denoised by an edge-preserving regularization method. Extensive experimental results indicate that our method performances better than many existing filters do for its robust image restoration and accurate noise detection. [ABSTRACT FROM AUTHOR]

Salinity in soils is among the major constraints of agricultural and forestry productivity worldwide. To investigate the basis of salt tolerance in Salix matsudana Koidz (Chinese willow), comparative proteomes of 1.5-month-old cuttings were analyzed using two-dimensional gel electrophoresis (2-DE). Proteins were extracted from roots of plants grown under 100 mM NaCl stress for 6, 24, or 48 h; whereas proteins from stems and leaves were extracted from plants grown under 100 mM NaCl stress for a period of 48 h. In total, 42, 17 and 19 differential protein spots were identified by MALDI-TOF/TOF in roots, stems and leaves, respectively. The majority of these salt stress-responsive proteins were organ specific. The identified proteins were involved in 12 metabolic pathways and processes, and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) analysis revealed that some of the proteins, such as USP-like, PR-10 and quinone reductase, were regulated by salt stress at the level of transcription. Pathogenesis-related proteins, in particular, were found to be abundant and up-regulated by salt stress. This suggests that PRs may be closely related to S. matsudana defense against salt stress, possibly by enhancing the resistance to opportunistic fungal attack. Based on the predicted functions of the identified proteins, we propose a salt stress response model describing a possible management strategy for cellular activities in S. matsudana under short-term salinity stress. Furthermore, some of the identified proteins might be potential candidates to improve salt tolerance in woody plants. [ABSTRACT FROM AUTHOR]

This paper presents an efficient random-valued impulse noise removal algorithm. The filtering process contains two phases: a detection phase followed by a filtering phase. In the detection phase, the proposed method uses the novel image statistics, the spatial local outlier measure (SLOM) and the Q-estimate, to identify impulses in a corrupted image. When the noise pixels are identified, their values are restored by an edge-preserving regularized method in the filtering phase. Extensive experimental results show that our filter provides a significant improvement over many other existing techniques. [ABSTRACT FROM AUTHOR]

Soil salinity is a serious problem worldwide. It is necessary to improve the salt tolerance of plants to avoid the progressive deterioration of saline soil. We showed that the over-expression of AtNHX1 improves salt tolerance in a transgenic poplar ( Populus deltoides CL × P. euramericana CL 'NL895') under mannose selection. Four transgenic poplar plants were obtained. Southern blot analysis showed that the pmi gene had integrated into the genome of the poplar. RT-PCR confirmed that AtNHX1 could be expressed normally in the transgenic plants. When tested for salt tolerance by NaCl stress, we measured a 100% increase in Na content in the three transgenic lines (T18, T50, T98) significantly higher than the 33% increase seen in wild-type plants. The chlorophyll content of the transgenic plants was not altered significantly, while the chlorophyll content in the control plants showed a small decrease. MDA content was decreased in the transgenic plants. These results show that the AtNHX1 gene may enhance salt tolerance due to increased vacuolar compartmentalization of sodium ions. [ABSTRACT FROM AUTHOR]

MicroRNAs (miRNAs) are highly conserved, endogenous, non-coding small RNAs of about 18-24 nucleotides in length. As negative regulators, miRNAs inhibit gene expressions at the post-transcriptional level by guiding cleavage or attenuating the translation of target genes, which affects organismal morphogenesis, development, and adaptability to changes occurring in the environment. Although miRNAs have been found in Populus species and have been shown to have diverse biological functions, few have been described in Populus cathayana Rehd. In this study, we have identified novel miRNAs from P. cathayana by constructing a library of small (18-24 nucleotides) RNA. After sequencing 691 small RNAs from P. cathayana, alignment to the Populus sequence database and secondary structure prediction were performed. These bioinformatic analyses identified five candidate miRNAs that were designated ptc-miR801, ptc-miR74, ptc-miR102, ptc-miR367, and ptc-miR213. Expression of these miRNAs was validated using real-time quantitative PCR analysis, and four of them displayed distinct expression patterns in leaves, stems, and roots. Our results show that four of the miRNAs are likely specific to leaves. Putative targets for these miRNAs play important roles in development and response to abiotic stress. [ABSTRACT FROM AUTHOR]

Plaque erosion (PE) with secondary thrombosis is one of the key mechanisms of acute coronary syndrome (ACS) which often leads to drastic cardiovascular events. Identification and prediction of PE are of fundamental significance for disease diagnosis, prevention and treatment. In vivo optical coherence tomography (OCT) data of eight eroded plaques and eight non-eroded plaques were acquired to construct three-dimensional fluid–structure interaction models and obtain plaque biomechanical conditions for investigation. Plaque stenosis severity, plaque burden, plaque wall stress (PWS) and strain (PWSn), flow shear stress (FSS), and ΔFSS (FSS variation in time) were extracted for comparison and prediction. A logistic regression model was used to predict plaque erosion. Our results indicated that the combination of mean PWS and mean ΔFSS gave best prediction (AUC = 0.866, 90% confidence interval (0.717, 1.0)). The best single predictor was max ΔFSS (AUC = 0.819, 90% confidence interval (0.624, 1.0)). The average of maximum FSS values from eroded plaques was 76% higher than that from the non-eroded plaques (127.96 vs. 72.69 dyn/cm2) while the average of mean FSS from erosion sites of the eight eroded plaques was 48.6% higher than that from sites without erosion (71.52 vs. 48.11 dyn/cm2). The average of mean PWS from plaques with erosion was 22.83% lower than that for plaques without erosion (83.2 kPa vs. 107.8 kPa). This pilot study suggested that combining plaque stress, strain and flow shear stress could help better identify patients with potential plaque erosion, enabling possible early intervention therapy. Further studies are needed to validate our findings. [ABSTRACT FROM AUTHOR]