Search

BACKGROUND Enabling a health system to learn from its historical and emerging data is a primary focus of medical AI research. Though digital pathology (DP) hasn’t gained similar popularity as clinical radiology and hospitalization research, its semistructured data drove natural language processing (NLP) to reveal codable insights from textual data. However, obtaining high-quality annotated samples as yet depended on predefined templates or human annotators, which became a bottleneck of automation. We noticed the prolonged undermining of morphology electronic health records (EHR) and its potential to supply high-quality labels and be the stepping stone towards automatic AI and a self-learning system. OBJECTIVE To develop an annotation-free NLP pipeline with proper human control for auto-deriving precise codes that had been annotated by a health system’s pathologists, text preprocessing, constructing machine learning classifiers to annotate text with clinically precise codes, and enabling system-wide application of the designed NLP pipeline to investigate historical data and enhance health information, promotion, and communication. METHODS Using colorectal dysplasia as an example, we developed the NLP pipeline with EHR of a population who attended baseline colorectal procedures in Hong Kong’s public health system between 2000 and 2018 when aged 50-75 years. The high-quality morphology codes were precisely-graded dysplasia, where high-grade dysplasia served as the positive label. After identifying precisely-coded, ambiguously-coded, and unlabeled cases from the EHR, we standardized the textual data before feeding them into a bidirectional long short-term memory neural network classification model. Our experimental design examined factors including two kinds of the unit of text analysis (report-/episode-based), the active learning with text curation, and the minimum sample size required for training an accurate classifier. Model performance was measured in testing and validation sets by the area under the receiver operating curve (AUC). RESULTS More than 35% of eligible text reports mentioned dysplasia. Precisely-graded dysplasia yielded a low prevalence in morphology codes. Still, the NLP pipeline identified more than 10,000 cases of high-grade dysplasia, which supplied a sufficient amount of positive cases for proving the efficacy of the proposed NLP pipeline. All testing AUCs of report-based active learning with text curation exceeded 0.88. A 200 sample size or more could secure 0.95 testing AUCs with active learning of text curation. Holding other factors the same, validation AUCs were worse than testing AUCs, indicating ambiguously-labeled cases were likely of the complex original text. CONCLUSIONS We demonstrated the feasibility, novel performances, and applications in automating annotation-free NLP pipelines at a system level. Our interdisciplinary pipeline can be a formal standard approach for a health system to realize self-learning from semistructured pathology EHR, with an orientation of precision public health and better person-centric care.

18S, 5.8S, and 28S ribosomal RNAs (rRNAs) are cotranscribed as a pre-ribosomal RNA (pre-rRNA) from the rDNA by RNA polymerase I whose activity is vigorous during the S-phase, leading to a conflict with rDNA replication. This conflict is resolved partly by replication-fork-barrier (RFB)-sites sequences located downstream of the rDNA and RFB-binding proteins such as Ttf1. However, how Ttf1 is displaced from RFB-sites to allow replication fork progression remains elusive. Here, we reported that loss-of-function of Bms1l, a nucleolar GTPase, upregulates rDNA transcription, causes replication-fork stall, and arrests cell cycle at the S-to-G2 transition; however, the G1-to-S transition is constitutively active characterized by persisting DNA synthesis. Concomitantly, ubf, tif-IA, and taf1b marking rDNA transcription, Chk2, Rad51, and p53 marking DNA-damage response, and Rpa2, PCNA, Fen1, and Ttf1 marking replication fork stall are all highly elevated in bms1l mutants. We found that Bms1 interacts with Ttf1 in addition to Rc1l. Finally, we identified RFB-sites for zebrafish Ttf1 through chromatin immunoprecipitation sequencing and showed that Bms1 disassociates the Ttf1‒RFB complex with its GTPase activity. We propose that Bms1 functions to balance rDNA transcription and replication at the S-phase through interaction with Rcl1 and Ttf1, respectively. TTF1 and Bms1 together might impose an S-phase checkpoint at the rDNA loci.

Forest ecosystems are subject to recurring fires as one of their most significant disturbances. Accurate mapping of burn severity is crucial for post-fire land management and vegetation regeneration monitoring. Remote-sensing-based monitoring of burn severity faces new challenges when forests experience both fire and non-fire disturbances, which may change the biophysical and biochemical properties of trees in similar ways. In this study, we develop a Disturbance Weighting Analysis Model (DWAM) for accurately mapping burn severity in a forest landscape that is jointly affected by wildfire and an emerging infectious disease – sudden oak death. Our approach treats burn severity in each basic mapping unit (e.g., 30 m grid from a post-fire Landsat image) as a linear combination of burn severity of trees affected (diseased) and not affected by the disease (healthy), weighted by their areal fractions in the unit. DWAM is calibrated using two types of inputs: i) look-up tables (LUTs) linking burn severity and post-fire spectra for diseased and healthy trees, derived from field observations, hyperspectral sensors [e.g., Airborne Visible InfraRed Imaging Spectrometer (AVIRIS)], and radiative transfer models; and ii) pre-fire fractional maps of diseased and healthy trees, derived by decomposing a pre-fire Landsat image using Multiple Endmember Spectral Mixture Analysis (MESMA). Considering the presence of tree disease in DWAM improved the overall map accuracy by 42%. The superior performance is consistent across all three stages of disease progression. Our approach demonstrates the potential for improved mapping of forest burn severity by reducing the confounding effects of other biotic disturbances.

Relatively high abundance of 17α (H)-diahopane has been detected in source rock extracts in the Yabulai Basin, a lacustrine basin in northwest China. Gas chromatography-mass spectrometry (GC-MS), X-ray diffraction (XRD) and Field emission environment scanning electron microscope imaging (FE-ESEM) were used in combination with conventional geochemical parameters to unravel the enrichment mechanisms. Significant variations of 17α (H)-diahopane concentrations detected in the Xinhe Formation with similar maturity inferred that thermal evolution is not the unique factor inducing the compound generating. The sedimentation of samples rich in 17α (H)-diahopane usually occur under suboxic conditions. Results of mineralogy analysis suggested that samples with high content of 17α (H)-diahopane are rich in carbonate minerals (about 20 wt %), which might deposit in brackish-saline water depositional environment. Relatively high content of gammacerane is also consistent with the experimental results of XRD. Based on the viewpoint of Moldowan et al. (1991), the regular hopanes may be the precursor of the rearranged hopanes, with the Wagner-Meerwein rearrangement as the mechanism. In this reaction, the critical factor of methyl migration is the production of carbocation starting with dehydroxylation. We suggested a general scheme for the generation of 17α (H)-diahopane in this study. In lacustrine sediments, suboxic and brackish environment could catalyze the methyl migration accompanied by carbonate deposit. Its geochemical implication can be used to reconstruct lacustrine paleoenvironment and contribute to a better understanding on oil-oil correlation, oil-source rock correlation and petroleum system study.

Road scene parsing is a common task in semantic segmentation. Its images have characteristics of containing complex scene context and differing greatly among targets of the same category from different scales. To address these problems, we propose a semantic segmentation model combined with edge detection. We extend the segmentation network with an encoder-decoder structure by adding an edge feature pyramid module, namely Edge Feature Pyramid Network (EFPNet, for short). This module uses edge detection operators to get boundary information and then combines the multiscale features to improve the ability to recognize small targets. EFPNet can make up the shortcomings of convolutional neural network features, and it helps to produce smooth segmentation. After extracting features of the encoder and decoder, EFPNet uses Euclidean distance to compare the similarity between the presentation of the encoder and the decoder, which can increase the decoder’s ability to restore from the encoder. We evaluated the proposed method on Cityscapes datasets. The experiment on Cityscapes datasets demonstrates that the accuracies are improved by 7.5% and 6.2% over the popular SegNet and ENet. And the ablation experiment validates the effectiveness of our method.

The rapid progress of photorealistic synthesis techniques has reached at a critical point where the boundary between real and manipulated images starts to blur. Thus, benchmarking and advancing digital forgery analysis have become a pressing issue. However, existing face forgery datasets either have limited diversity or only support coarse-grained analysis. To counter this emerging threat, we construct the ForgeryNet dataset, an extremely large face forgery dataset with unified annotations in image- and video-level data across four tasks: 1) Image Forgery Classification, including two-way (real / fake), three-way (real / fake with identity-replaced forgery approaches / fake with identity-remained forgery approaches), and n-way (real and 15 respective forgery approaches) classification. 2) Spatial Forgery Localization, which segments the manipulated area of fake images compared to their corresponding source real images. 3) Video Forgery Classification, which re-defines the video-level forgery classification with manipulated frames in random positions. This task is important because attackers in real world are free to manipulate any target frame. and 4) Temporal Forgery Localization, to localize the temporal segments which are manipulated. ForgeryNet is by far the largest publicly available deep face forgery dataset in terms of data-scale (2.9 million images, 221,247 videos), manipulations (7 image-level approaches, 8 video-level approaches), perturbations (36 independent and more mixed perturbations) and annotations (6.3 million classification labels, 2.9 million manipulated area annotations and 221,247 temporal forgery segment labels). We perform extensive benchmarking and studies of existing face forensics methods and obtain several valuable observations., Comment: 17 pages, 11 figures, Accepted to CVPR 2021 (Oral), project webpage: https://yinanhe.github.io/projects/forgerynet.html

Quality assurance and quality control (QA&QC) procedures are vital for ensuring data reliability, but little is known about the use of such procedures in reducing airborne microplastic (MP) contamination. To address this issue, we tried to determine the efficiency of two common methods (washing and ashing experimental glassware) for removing airborne MPs and identified airborne contamination during MP analytical procedure. The results showed the removal efficiencies of washing and ashing were an average of 88%-98% and 100%, respectively, indicating that both methods could eliminate most of the spiked airborne MPs with no significant difference noted between the two methods. Although rigorous measures were taken to prevent contamination from ambient air, trace amounts of airborne MPs were still detected, which is an issue that has not been adequately investigated in previous studies. All of the procedural contaminants detected in this study were fibrous. Approximately 88% of these fibers were cotton-like (cotton, cellulose, and cellophane) fibers, and 13% of them were plastic. Surprisingly, cotton-like fibers and MPs had a similar size distribution, suggesting that they may have undergone a similar weathering process. In the end, to cope with inevitable airborne contamination, several measures were proposed for further research. Such measures will provide the necessary methodological assistance for accurate quantification of MP pollution in the field.

Over the last two decades (since ca. 2000), Geographic Object-Based Image Analysis (GEOBIA) has emerged as a new paradigm to analyzing high-spatial resolution remote-sensing imagery. During this ti...

After metal recovery, the non-metallic fraction of waste printed circuit board (WPCB) could still be valuable. It is rich in organic polymer and inorganics which could be a source of chemicals and fuel. In this study, WPCB was pyrolyzed under different conditions and solid, liquid and gaseous products were collected and characterized in details. The resultant solid products were powdered, low in porosity but rich in surface functionals groups. Organics in original WPCB were charred and glass fibers was able to maintained its cylindrical configuration. Pyrolysis at 550 °C for 1 h led to the highest yield of liquid which was predominated with phenol and alkylated phenols. Brominated organics were also found in oil though their presence and composition varied with pyrolysis temperature. Gaseous products were mainly composed of CO2, CH4 and H2. Based on the compositions of pyrolysis products, pathways of WPCB disintegration were proposed.

Background Many countries/regions implemented strict border measures (e.g., 14-day quarantines) as a blanket policy to prevent COVID-19 importations, while proposed "travel bubbles" as an alternative to reduce the impact of border controls. We aim to examine the differential importation risks with departure origins and post-arrival controls. Methods We developed a Bayesian framework to model disease progress of COVID-19 and the effectiveness of travel measures and inferred the origin-specific disease prevalence among inbound travellers, using data on passengers arriving in Hong Kong and laboratory-confirmed imported cases. We estimated the origin-specific risks of releasing infectious travellers under different control strategies and traveller volumes. We also estimated the risk of having released infectious travellers when a resurgence occurs in departure locations with no imported cases during a certain period. Findings Under the then strict controls of 14-day quarantine and testing on day 12, the Philippines imposed the greatest importation risk among the studied countries/regions (95.8% of releasing at least one infectious traveller, 95% credible interval (CrI), 94.8-96.6%). This was higher than that from low prevalence countries/regions (e.g., 23.4%, 95% CrI, 21.6-25.3% for Taiwan) if controls relaxed (i.e., 7-day quarantine and test on day 5). Increased traveller volumes and resurgence in departure locations with low prevalence under relaxed controls did not impose a greater importation risk than high prevalence locations under stricter controls. Interpretation Moderate relaxation of control measures for travellers arriving from low prevalence locations did not impose higher risks of community outbreaks than strict controls on travellers from high prevalence locations. Funding Health and Medical Research Fund, Hong Kong.

Environmental disturbance regimes are more frequently being altered by historically novel events and disturbance interactions, which may trigger reorganizations of new ecosystem states and processes. Here we examine synergies between emerging forest disease and wildfire to determine whether disease outbreak changes environmental drivers of burn severity using sudden oak death and the basin complex fire in California as a case study of novel disturbance interaction. We mapped the spatial distribution of sudden oak death tree mortality using a new object-based filter with 1.0 m resolution KOMPSAT-2 images. We integrated these data with a physical simulation model of burn severity informed by post-fire Landsat data. Model performance varied across stages of disease establishment (early, middle and late) with stronger relationships occurring during later stages of disease progression. Multiscale statistical analysis of environmental drivers of burn severity in diseased compared to healthy forests showed that sudden oak death tree mortality altered relationships between burn severity and the biophysical environment. Specifically, compared to the healthy forests, those affected by disease exhibited higher landscape heterogeneity at smaller spatial scales (e.g., 25 and 50 m), which has been associated with decreased burn severity in the literature. Our results showed the opposite pattern. That is, a disease-affected landscape comprising less connected patches and higher patch shape complexity was more likely to experience greater burn severity. This suggests that disease-caused increases in surface fuels may have reduced the landscape's resistance to fire and in turn increased burn severity in forest patches neighboring disease-impacted forests.

The detection of Near-Duplicate Video has been studied by many scholars in recent years. This paper proposes a method for classifying and identifying infringing video by extracting video visual features, obtaining inter-frame relationship graphs, and using CNN classification network. We apply the improved weighting technique of text mining to video solution and still achieve better results. This method is robust in experiments and does not require retraining of a model in practical applications. At the same time, this method can also be applied to large-scale video retrieval. Our method has an F1 score of 95.74%.

Forest ecosystems are increasingly affected by a range of tree mortality events, which may permanently alter forest functional traits and disrupt their ecosystem services. While individual forest disturbances are well studied, interactions between multiple disturbances and changes of spatial patterns of forested landscapes are rarely quantified. In this study, we aim to analyze the role of wildfire in the Big Sur ecoregion of California on the spread of Phytophthora ramorum, an invasive pathogen which causes sudden oak death, the most important driver of mortality across 1000 km of coastal, fire-prone mixed conifer, evergreen hardwood, and woodlands. We investigated two questions specific to the impacts of these disturbances at the landscape scale: (i) did rates of P. ramorum caused tree mortality change after wildfire? (ii) Following the wildfire, to what degree did the continued disease-driven mortality alter forest distribution? To answer these questions, we analyzed remote-sensing-derived products of post-fire burn severity and maps of disease-driven tree mortality. Quantification of burn severity and post fire disease mortality for the burned and unburned areas provided reference conditions for statistical hypothesis tests. The results from statistical and three landscape pattern analyses (area, shape, and isolation) suggest a significant role of wildfire in the reemergence of this invasive pathogen. First, rates of disease caused mortality after wildfire was negatively associated with burn severity suggesting some fire-driven containment of disease during post-fire forest recovery. Second, disease was positively correlated with the distance to fire boundary in unburned areas suggesting the effects of fire on disease extended into unburned areas while attenuating with distance from the burn. Lastly, wildfire reduced area, edge density and isolation of healthy tree patches and these effects did not recover to pre-fire levels for any of the three metrics after eight years of vegetation recovery. Given the widespread prevalence of disease-driven mortality, the importance and frequency of fire, as well as the naturalization of Phytophthora ramorum across a broad geographic area, these fire-disease interactions have potential to shape forest structure and disease dynamics across millions of acres of forested wildlands in California and Oregon.

Forest ecosystems have been increasingly affected by a variety of disturbances, including emerging infectious diseases (EIDs), causing extensive tree mortality in the Western United States. Especially over the past decade, EID outbreaks occurred more frequently and severely in forest landscapes, which have killed large numbers of trees. While tree mortality is observable from remote sensing, its symptom may be associated with both disease and non-disease disturbances (e.g., wildfire and drought). Species distribution modeling is widely used to understand species spatial preferences for certain habitat conditions, which may constrain uncertain remote sensing approaches due to limited spatial and spectral resolution. In this study, we integrated multi-sensor remote sensing and species distribution modeling to map disease-caused tree mortality in a forested area of 80,000 ha from 2005 to 2016. We selected sudden oak death (caused by pathogen P. ramorum) as a case study of a rapidly spreading emerging infectious disease, which has killed millions of oak (Quercus spp.) and tanoak (Lithocarpus densiflorus) in California over the past decades. To balance the needs for fine-scale monitoring of disease distribution patterns and satisfactory coverage at broad scales, our method applied spectral unmixing to extract sub-pixel disease presence using yearly Landsat time series. The results were improved by employing the probability of disease infection generated from a species distribution model. We calibrated and validated the method with image samples from high-spatial resolution NAIP (National Agriculture Imagery Program), and hyperspectral AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) sensors, Google Earth® imagery, and field observations. The findings reveal an annual sudden oak death infection rate of 7% from 2005 to 2016, with overall mapping accuracies ranging from 76% to 83%. The integration of multi-sensor remote sensing and species distribution modeling considerably reduced the overestimation of disease effects as compared to the use of remote sensing alone, leading to an average of 26% decrease in detecting disease-affected trees. Such integration strategy proved the effectiveness of mapping long-term, disease-caused tree mortality in forest landscapes that have experienced multiple disturbances.

To provide data for the occupational health supervision by analyzing the occupational health status in manufacturing industry in Guangzhou, China.The occupational health investigation was performed in 280 enterprises randomly selected from 8 industries based on industry stratification. According to the occupational health standards, 198 out of the 280 enterprises were supervised and monitored. Sample testing was performed in 3~5 workplaces where workers were exposed to the highest concentration/intensity of occupational hazard for the longest time. Comparative analyses of the overproof rates of hazard were performed among enterprises, workplaces, and testing items from different industries.The concentrations of occupational hazard in 42.93% (85/198) of enterprises and 22.96% (200/871) of workplaces were above the limit concentration. The most severe hazards were the noises in shipbuilding and wooden furniture industries and the welding fumes in shipbuilding industry. Less than 30% of enterprises were able to provide occupational health examination and periodic test reports of occupational hazard in workplaces. The rate of the workers with abnormal occupational health examination results and the need for reexamination reached 6.63% (832/12 549), and they were mostly from shipbuilding, wooden furniture, and chemical industries.The occupational health supervision should be strengthened in enterprises, and hazard from noises and dusts should be selectively controlled or reduced. The publication of relevant data and information of occupational health in enterprises should be promoted to enhance social supervision.

The article argues that the harmonisation of national memories facilitates genuine reconciliation, while memory divergence resulting from national mythmaking hampers reconciliation. After World War II, Sino-Japanese and West German–Polish relations were antagonised by the Cold War structure, and pernicious myths prevailed in national collective memory. Then China and Japan brushed aside historical legacy for immediate diplomatic normalisation, but their reconciliation was impeded by elite mythmaking practices. Since the 1970s West Germany and Poland have de-mythified war history and engaged in historical settlement, paving the way for deep reconciliation after the Cold War.

After nearly sixty years of political confrontation, hopes for cross-Taiwan Strait reconciliation have run high since the traditionally pro-unification Nationalist Party (the Kuomintang, KMT) returned to power in Taiwan in May 2008. However, obstacles to reconciliation remain daunting, due to a fundamental disjuncture between the ideological beliefs of the two sides, in particular because China and Taiwan still lack a shared memory of Taiwanese history that can serve as the foundation for their reconciliation. This article examines a wide variety of sources from Taiwan and China over recent decades. It illustrates their conspicuous memory gap over the history of the island. Cross-Strait reconciliation needs to begin with recognizing rather than ignoring or covering up the memory gap. Dialogue and joint studies should be carried out to better understand each other’s political perspective and emotional appeal associated with historical memory.

Sino-Japanese political relations, fraught with disputes and tension during the Koizumi years, only began to recover after Abe came to power. This article investigates the driving forces shaping recent and future bilateral relations. Using evidence from the Koizumi era, I argue that 1) bilateral commercial links prove a weak stabilizing factor for political relations; 2) the current distribution of power between China and Japan does not dictate their strategic rivalry, but they may still treat each other as rivals if they perceive the danger of long-term power transition and mutual hostile intent; 3) the frequent flare-up of bilateral history disputes can exacerbate mutual threat perceptions among elites and generate popular emotional pressure for hard-line government policy toward the other country. The future of Sino-Japanese cooperation heavily depends on their efforts to resolve the negative historical legacy.

Anti-Japanese popular nationalism is rising high in China today. Little evidence to date proves that it is officially orchestrated. Nonetheless, Chinese popular nationalism still has deep roots in the state's history propaganda which has implanted pernicious myths in the national collective memory. Fueling mistrust and exacerbating a mutual threat perception, popular nationalism could be a catalyst for future Sino–Japanese conflict over the Taiwan problem, island disputes, and maritime resource competition. The increasingly liberalized but often biased Chinese media, the role of nationalist sub-elites, and the government's accommodation have all contributed to the strength of anti-Japanese nationalism, which cannot be mitigated by bilateral economic interdependence. To rid bilateral relations of the negative historical legacy, the two countries need the vision and determination to remove nationalistic myths and promote a shared history through mutual critique and self-reflection in transnational historian...

East Asia is a promising setting for testing and refining IR theories, and the decade of the 2000s is an ideal time to do so. Although prior to developments in the 2010s interest in the significance of East Asia for IR theory was surprisingly limited, during the 2000s it was already leading a transition away from the order many had assumed was replacing the Cold War toward a new regional order. Within the region, bilateral relations are of particular interest for grasping the changes under way. Focusing on three of them, this chapter raises fundamental questions about the prevailing IR theory.