The Chinese University of Hong Kong

About: The Chinese University of Hong Kong is a education organization based out in Hong Kong, China. It is known for research contribution in the topics: Population & Cancer. The organization has 43411 authors who have published 93672 publications receiving 3066651 citations.

Hierarchical Image Saliency Detection on Extended CSSD

Abstract: Complex structures commonly exist in natural images. When an image contains small-scale high-contrast patterns either in the background or foreground, saliency detection could be adversely affected, resulting erroneous and non-uniform saliency assignment. The issue forms a fundamental challenge for prior methods. We tackle it from a scale point of view and propose a multi-layer approach to analyze saliency cues. Different from varying patch sizes or downsizing images, we measure region-based scales. The final saliency values are inferred optimally combining all the saliency cues in different scales using hierarchical inference. Through our inference model, single-scale information is selected to obtain a saliency map. Our method improves detection quality on many images that cannot be handled well traditionally. We also construct an extended Complex Scene Saliency Dataset (ECSSD) to include complex but general natural images.

Viral Loads and Duration of Viral Shedding in Adult Patients Hospitalized with Influenza

Abstract: 5.06 � 1.85 3.62 � 2.13 P p .005 +0.03 to +1.68)). Viral RNA concentration demonstrated a nonlinear decrease with time; 26% of oseltamivir- treated and 57% of untreated patients had RNA detected at 1 week after symptom onset. Oseltamivir started on or before symptom day 4 was independently associated with an accelerated decrease in viral RNA concentration (mean b (standard error), 1.19 (0.43) and 0.68 (0.33) log10 copies/mL for patients treated on day 1 and days 2-3, respectively; ) and viral RNA clearance at 1 week (odds ratio, 0.10 (95% confidence interval, 0.03- P ! .05 0.35) and 0.30 (0.10-0.90) for patients treated on day 1-2 and day 3-4, respectively). Conversely, major comor- bidities and systemic corticosteroid use for asthma or chronic obstructive pulmonary disease exacerbations were associated with slower viral clearance. Viral RNA clearance was associated with a shorter hospital stay (7.0 vs 13.5 days; ). P p .001 Conclusion. Patients hospitalized with severe influenza have more active and prolonged viral replication. Weakened host defenses slow viral clearance, whereas antivirals started within the first 4 days of illness enhance viral clearance.

Gut mucosal microbiome across stages of colorectal carcinogenesis

Abstract: Gut microbial dysbiosis contributes to the development of colorectal cancer (CRC). Here we catalogue the microbial communities in human gut mucosae at different stages of colorectal tumorigenesis. We analyse the gut mucosal microbiome of 47 paired samples of adenoma and adenoma-adjacent mucosae, 52 paired samples of carcinoma and carcinoma-adjacent mucosae and 61 healthy controls. Probabilistic partitioning of relative abundance profiles reveals that a metacommunity predominated by members of the oral microbiome is primarily associated with CRC. Analysis of paired samples shows differences in community configurations between lesions and the adjacent mucosae. Correlations of bacterial taxa indicate early signs of dysbiosis in adenoma, and co-exclusive relationships are subsequently more common in cancer. We validate these alterations in CRC-associated microbiome by comparison with two previously published data sets. Our results suggest that a taxonomically defined microbial consortium is implicated in the development of CRC.

Prevalence of amyloid PET positivity in dementia syndromes: a meta-analysis.

Abstract: Importance Amyloid-β positron emission tomography (PET) imaging allows in vivo detection of fibrillar plaques, a core neuropathological feature of Alzheimer disease (AD). Its diagnostic utility is still unclear because amyloid plaques also occur in patients with non–AD dementia. Objective To use individual participant data meta-analysis to estimate the prevalence of amyloid positivity on PET in a wide variety of dementia syndromes. Data Sources The MEDLINE and Web of Science databases were searched from January 2004 to April 2015 for amyloid PET studies. Study Selection Case reports and studies on neurological or psychiatric diseases other than dementia were excluded. Corresponding authors of eligible cohorts were invited to provide individual participant data. Data Extraction and Synthesis Data were provided for 1359 participants with clinically diagnosed AD and 538 participants with non–AD dementia. The reference groups were 1849 healthy control participants (with amyloid PET) and an independent sample of 1369 AD participants (with autopsy data). Main Outcomes and Measures Estimated prevalence of positive amyloid PET scans according to diagnosis, age, and apolipoprotein E (APOE) e4 status, using the generalized estimating equations method. Results The likelihood of amyloid positivity was associated with age and APOE e4 status. In AD dementia, the prevalence of amyloid positivity decreased from age 50 to 90 years in APOE e4 noncarriers (86% [95% CI, 73%-94%] at 50 years to 68% [95% CI, 57%-77%] at 90 years; n = 377) and to a lesser degree in APOE e4 carriers (97% [95% CI, 92%-99%] at 50 years to 90% [95% CI, 83%-94%] at 90 years; n = 593; P Conclusions and Relevance Among participants with dementia, the prevalence of amyloid positivity was associated with clinical diagnosis, age, and APOE genotype. These findings indicate the potential clinical utility of amyloid imaging for differential diagnosis in early-onset dementia and to support the clinical diagnosis of participants with AD dementia and noncarrier APOE e4 status who are older than 70 years.

Scene Graph Generation from Objects, Phrases and Region Captions

Abstract: Object detection, scene graph generation and region captioning, which are three scene understanding tasks at different semantic levels, are tied together: scene graphs are generated on top of objects detected in an image with their pairwise relationship predicted, while region captioning gives a language description of the objects, their attributes, relations and other context information. In this work, to leverage the mutual connections across semantic levels, we propose a novel neural network model, termed as Multi-level Scene Description Network (denoted as MSDN), to solve the three vision tasks jointly in an end-to-end manner. Object, phrase, and caption regions are first aligned with a dynamic graph based on their spatial and semantic connections. Then a feature refining structure is used to pass messages across the three levels of semantic tasks through the graph. We benchmark the learned model on three tasks, and show the joint learning across three tasks with our proposed method can bring mutual improvements over previous models. Particularly, on the scene graph generation task, our proposed method outperforms the stateof- art method with more than 3% margin. Code has been made publicly available.