Cross-modal retrieval aims to enable flexible retrieval experience across different modalities (e.g., texts vs. images). The core of crossmodal retrieval research is to learn a common subspace where the items of different modalities can be directly compared to each other. In this paper, we present a novel Adversarial Cross-Modal Retrieval (ACMR) method, which seeks an effective common subspace based on adversarial learning. Adversarial learning is implemented as an interplay between two processes. The first process, a feature projector, tries to generate a modality-invariant representation in the common subspace and to confuse the other process, modality classifier, which tries to discriminate between different modalities based on the generated representation. We further impose triplet constraints on the feature projector in order to minimize the gap among the representations of all items from different modalities with same semantic labels, while maximizing the distances among semantically different images and texts. Through the joint exploitation of the above, the underlying cross-modal semantic structure of multimedia data is better preserved when this data is projected into the common subspace. Comprehensive experimental results on four widely used benchmark datasets show that the proposed ACMR method is superior in learning effective subspace representation and that it significantly outperforms the state-of-the-art cross-modal retrieval methods. CCS CONCEPTS• Information systems → Multimedia and multimodal retrieval;

BackgroundSeveral systematic reviews and meta-analyses demonstrated the association between depression and the risk of coronary heart disease (CHD), but the previous reviews had some limitations. Moreover, a number of additional studies have been published since the publication of these reviews. We conducted an updated meta-analysis of prospective studies to assess the association between depression and the risk of CHD.MethodsRelevant prospective studies investigating the association between depression and CHD were retrieved from the PubMed, Embase, Web of Science search (up to April 2014) and from reviewing reference lists of obtained articles. Either a random-effects model or fixed-effects model was used to compute the pooled risk estimates when appropriate.ResultsThirty prospective cohort studies with 40 independent reports met the inclusion criteria. These groups included 893,850 participants (59,062 CHD cases) during a follow-up duration ranging from 2 to 37 years. The pooled relative risks (RRs) were 1.30 (95% CI, 1.22-1.40) for CHD and 1.30 (95% CI, 1.18-1.44) for myocardial infarction (MI). In the subgroup analysis by follow-up duration, the RR of CHD was 1.36 (95% CI, 1.24-1.49) for less than 15 years follow-up, and 1.09 (95% CI, 0.96-1.23) for equal to or more than 15 years follow-up. Potential publication bias may exist, but correction for this bias using trim-and-fill method did not alter the combined risk estimate substantially.ConclusionsThe results of our meta-analysis suggest that depression is independently associated with a significantly increased risk of CHD and MI, which may have implications for CHD etiological research and psychological medicine.Electronic supplementary materialThe online version of this article (doi:10.1186/s12888-014-0371-z) contains supplementary material, which is available to authorized users.

Background and objectives Comprehensive epidemiologic data on AKI are particularly lacking in Asian countries. This study sought to assess the epidemiology and clinical correlates of AKI among hospitalized adults in China.Design, setting, participants, & measurements This was a multicenter retrospective cohort study of 659,945 hospitalized adults from a wide range of clinical settings in nine regional central hospitals across China in 2013. AKI was defined and staged according to Kidney Disease Improving Global Outcomes criteria. The incidence of AKI in the cohort was estimated using a novel two-step approach with adjustment for the frequency of serum creatinine tests and other potential confounders. Risk factor profiles for hospital-acquired (HA) and communityacquired (CA) AKI were examined. The in-hospital outcomes of AKI, including mortality, renal recovery, length of stay, and daily cost, were assessed.Results The incidence of CA-AKI and HA-AKI was 2.5% and 9.1%, respectively, giving rise to an overall incidence of 11.6%. Although the risk profiles for CA-AKI and HA-AKI differed, preexisting CKD was a major risk factor for both, contributing to 20% of risk in CA-AKI and 12% of risk in HA-AKI. About 40% of AKI cases were possibly drug-related and 16% may have been induced by Chinese traditional medicines or remedies. The in-hospital mortality of AKI was 8.8%. The risk of in-hospital death was higher among patients with more severe AKI. Preexisting CKD and need for intensive care unit admission were associated with higher death risk in patients at any stage of AKI. Transiency of AKI did not modify the risk of in-hospital death. AKI was associated with longer length of stay and higher daily costs, even after adjustment for confounders.Conclusion AKI is common in hospitalized adults in China and is associated with significantly higher in-hospital mortality and resource utilization.