Ensemble Prototype Network For Weakly Supervised Temporal Action Localization.

Weakly supervised temporal action localization (TAL) aims to localize the action instances in untrimmed videos using only video-level action labels. Without snippet-level labels, this task should be hard to distinguish all snippets with accurate action/background categories. The main difficulties are the large variations brought by the unconstraint background snippets and multiple subactions in action snippets. The existing prototype model focuses on describing snippets by covering them with clusters (defined as prototypes). In this work, we argue that the clustered prototype covering snippets with simple variations still suffers from the misclassification of the snippets with large variations. We propose an ensemble prototype network (EPNet), which ensembles prototypes learned with consensus-aware clustering. The network stacks a consensus prototype learning (CPL) module and an ensemble snippet weight learning (ESWL) module as one stage and extends one stage to multiple stages in an ensemble learning way. The CPL module learns the consensus matrix by estimating the similarity of clustering labels between two successive clustering generations. The consensus matrix optimizes the clustering to learn consensus prototypes, which can predict the snippets with consensus labels. The ESWL module estimates the weights of the misclassified snippets using the snippet-level loss. The weights update the posterior probabilities of the snippets in the clustering to learn prototypes in the next stage. We use multiple stages to learn multiple prototypes, which can cover the snippets with large variations for accurate snippet classification. Extensive experiments show that our method achieves the state-of-the-art weakly supervised TAL methods on two benchmark datasets, that is, THUMOS'14, ActivityNet v1.2, and ActivityNet v1.3 datasets.

Active Factor Graph Network for Group Activity Recognition.

Group activity recognition aims to identify a consistent group activity from different actions performed by respective individuals. Most existing methods focus on learning the interaction between each two individuals (i.e., second-order interaction). In this work, we argue that the second-order interactive relation is insufficient to address this task. We propose a third-order active factor graph network, which models the third-order interaction in each pair of three active individuals. At first, to alleviate the noisy individual actions, we select active individuals by measuring each individual's influence. The individuals with the top-k largest influence weights are selected as active individuals. Then, for each three-individuals pair, we build a new factor node and contact the factor node with these individual nodes. In other words, we extend the base second-order interactive graph to a new third-order interactive graph, which is defined as factor graph. Next, we design a two-branch factor graph network, in which one branch is to consider all individuals (denoted as full factor graph) and the other one takes the active individuals into consideration (denoted as active factor graph). We leverage both the active and full factor graphs comprehensively for group activity recognition. Besides, to enforce group consistency, a consistency-aware reasoning module is designed with two penalty terms, which describe the inconsistency between individual actions and group activity respectively. Extensive experiments demonstrate that our method achieves state-of-the-art performance on four benchmark datasets, i.e., Volleyball, Collective Activity, Collective Activity Extended, and SoccerNet-v3 datasets. Visualization results further validate the interpretability of our method.

Subtraction improves the accuracy of coronary CT angiography for detecting obstructive disease in severely calcified segments.

OBJECTIVE: To investigate the accuracy, diagnostic confidence, and interobserver agreement of subtraction coronary CT angiography (CCTA) versus invasive coronary angiography on 320-row CT in coronary segments with severe or non-severe calcification. MATERIALS/METHODS: Sixty-four patients (33 men, 66.6 ± 8.2 years) with suspected coronary artery disease (CAD) were prospectively enrolled from October 2019 to June 2020. The cross-sectional circumferential extent of calcification was used to classify calcified segments as non-severely ( < 180°) or severely calcified ( ≥ 180°). Three independent, blinded radiologists evaluated the severity of coronary stenosis. Interobserver agreement was evaluated using Fleiss' kappa (κ). A multiple-reader multiple-case receiver operating characteristic (ROC) method was conducted, and diagnostic accuracy was measured using the mean areas under the ROC curves (AUCs), with ≥ 50% stenosis as a cut-off. Diagnostic confidence, diagnostic accuracy, and interobserver agreement were compared between CCTA with or without subtraction information in severely and non-severely calcified segments. RESULTS: In cases with severe calcification (51 patients, 146 segments), CCTA with subtraction information achieved better diagnostic accuracy (per-patient AUC: 0.73 vs 0.57, p = 0.03; per-segment AUC: 0.85 vs 0.62, p = 0.01), diagnostic confidence (3.7 vs 2.6, p < 0.001), and interobserver agreement (κ: 0.59 vs 0.30). Diagnostic accuracy (per-patient AUC: 0.81 vs 0.93, p = 0.30; per-patient AUC: 0.79 vs 0.82, p = 0.54) was not increased in cases with non-severe calcification (13 patients, 190 segments). CONCLUSIONS: CCTA with subtraction information achieved better diagnostic accuracy in cases of severe calcification (circumferential extent ≥ 180°). However, for non-severe calcification (circumferential extent < 180°), the effect of calcium subtraction was unclear, as it did not improve diagnostic accuracy. KEY POINTS: • Subtraction coronary CT angiography achieves better diagnostic accuracy, higher diagnostic confidence, and increased interobserver agreement for severe calcification (circumferential extent ≥ 180°). • Calcium subtraction does not improve the diagnostic accuracy of coronary CT angiography for calcification with a circumferential extent of < 180°.

Faceted Text Segmentation via Multitask Learning.

Text segmentation is a fundamental step in natural language processing (NLP) and information retrieval (IR) tasks. Most existing approaches do not explicitly take into account the facet information of documents for segmentation. Text segmentation and facet annotation are often addressed as separate problems, but they operate in a common input space. This article proposes FTS, which is a novel model for faceted text segmentation via multitask learning (MTL). FTS models faceted text segmentation as an MTL problem with text segmentation and facet annotation. This model employs the bidirectional long short-term memory (Bi-LSTM) network to learn the feature representation of sentences within a document. The feature representation is shared and adjusted with common parameters by MTL, which can help an optimization model to learn a better-shared and robust feature representation from text segmentation to facet annotation. Moreover, the text segmentation is modeled as a sequence tagging task using LSTM with a conditional random fields (CRFs) classification layer. Extensive experiments are conducted on five data sets from five domains: data structure, data mining, computer network, solid mechanics, and crystallography. The results indicate that the FTS model outperforms several highly cited and state-of-the-art approaches related to text segmentation and facet annotation.

Plasma Ephrin-A1 level in a cohort of diabetic retinopathy patients.

BACKGROUND: To determine plasma ephrin-A1 and VEGF165 levels in a cohort of diabetic retinopathy patients. METHODS: Plasma ephrin-A1 and VEGF165 levels in fifty-five subjects including 19 individuals without diabetes (non-DM), 16 patients with diabetes (DM) but without diabetic retinopathy, and 20 patients with diabetic retinopathy (DR), were determined by ELISA. Serum creatinine, total cholesterol, fasting blood glucose and HbA1c were also measured. One-way ANOVA, Kruskal-Wallis Test, Mann-Whitney U Test corrected by Bonferroni, Pearson Correlation Analysis and Spearman Correlation Coefficient Analysis were used for data analysis. RESULTS: Ephrin-A1 expression could be detected in human plasma with an average of 1.52 ± 0.43 (mean ± SEM) ng/ml. In DR subjects, the plasma ephrin-A1concentration was 3.63 ± 4.63 ng/ml, which was significantly higher than that of the other two groups (non-DM: 0.27 ± 0.13 ng/ml, DM: 0.35 ± 0.34 ng/ml). The expression of VEGF165 in human plasma was 34.00 ± 42.55 pg/ml, with no statistical difference among the three groups. There was no correlation between ephrin-A1 and VEGF165 in human plasma, but there was a correlation between plasma ephrin-A1 and duration of diabetes. CONCLUSIONS: Plasma ephrin-A1 was highly expressed in patients with diabetic retinopathy, and there was no difference of plasma VEGF165 expression in patients with diabetic retinopathy compared to the other two groups, suggesting that changes of plasma ephrin-A1 may be a more sensitive biomarker than plasma VEGF165 in detecting diabetic retinopathy.

Mechanical characteristics of medical grade UHMWPE under dynamic compression.

The mechanical properties of medical grade ultrahigh molecular weight polyethylene (UHMWPE) are critical for the safety and integrity of UHMWPE implantation. Accordingly, the mechanical features of UHMWPE are tested under repeated stress-controlled and strain-controlled compression at room temperature. Some important effect factors, such as stress rate, mean stress, stress amplitude, strain rate, mean strain, strain range and multiple load steps are further considered in detail. Results indicate that the lower stress rate causes the greater accumulated plastic strain and the accumulated plastic strain rate becomes increasingly lower with increasing number of cycles. The strain range and accumulated plastic strain rate decrease rapidly in the first stage, and then become almost steady during the second stage. Especially, the accumulated plastic strain rate per cycle for each case is less than 0.01 %/cycle after the initial 100 cycles. This means that the plastic strain accumulates very slowly and the shakedown behavior always occurs. Moreover, obvious cyclic softening and stress relaxation behaviors can be observed under cyclic strain-controlled compression during the first 50 cycles. This indicates that the accumulated plastic stain in the initial 100 cycles and the cyclic stress relaxation during the first 50 cycles should be assessed for the functionality of UHMWPE implantation.

Ultrathin Cu2O as an efficient inorganic hole transporting material for perovskite solar cells.

We demonstrate that ultrathin P-type Cu2O thin films fabricated by a facile thermal oxidation method can serve as a promising hole-transporting material in perovskite solar cells. Following a two-step method, inorganic-organic hybrid perovskite solar cells were fabricated and a power conversion efficiency of 11.0% was achieved. We found that the thickness and properties of Cu2O layers must be precisely tuned in order to achieve the optimal solar cell performance. The good performance of such perovskite solar cells can be attributed to the unique properties of ultrathin Cu2O, including high hole mobility, good energy level alignment with CH3NH3PbI3, and longer lifetime of photo-excited carriers. Combining the merits of low cost, facile synthesis, and high device performance, ultrathin Cu2O films fabricated via thermal oxidation hold promise for facilitating the developments of industrial-scale perovskite solar cells.

Gracillin induces apoptosis in HL60 human leukemic cell line via oxidative stress and cell cycle arrest of G1.

Gracillin, a kind of steroidal saponin isolated from the root bark of wild yam Dioscorea nipponica has been reported to exert antitumor activity. In the present study, we investigated the anticancer activity of gracillin against HL60 cells, and evaluated the possible mechanism involved in its antineoplastic action. The cell proliferation was evaluated by cell counting Kit-8 (CCK-8) assay, gracillin inhibited the growth of HL60 cells in a time- and concentration-dependent manner. Flow cytometry was used to analyze the cell cycle distribution whereas Annexin V-FITC/PI flow cytometry analysis was carried out to confirm apoptosis induced by gracillin, Our results demonstrated that gracillin could induce cell cycle arrest of G1 and apoptosis in HL60 cells. Furthermore, based on the biochemical methods, induction of oxidative stress by gracillin was indicated by increased the content of malondialdehyde (MDA), and decreased superoxide dismutase (SOD) activity. In addition, real time-PCR verified the expression of apoptosis-related genes, the mRNA level of Bcl-2 was decreased dramatically, while Bax was remarkably increased by gracillin. Taken together, gracillin could induce cell cycle arrest, oxidative stress, and apoptosis in HL60 cells, and has the potential to be developed as an antitumor agent.

Temperature-dependent excitonic photoluminescence of hybrid organometal halide perovskite films.

Organometal halide perovskites have recently attracted tremendous attention due to their potential for photovoltaic applications, and they are also considered as promising materials in light emitting and lasing devices. In this work, we investigated in detail the cryogenic steady state photoluminescence properties of a prototypical hybrid perovskite CH3NH3PbI3-xClx. The evolution of the characteristics of two excitonic peaks coincides with the structural phase transition around 160 K. Our results further revealed an exciton binding energy of 62.3 ± 8.9 meV and an optical phonon energy of 25.3 ± 5.2 meV, along with an abnormal blue-shift of the band gap in the high-temperature tetragonal phase.

Surface passivation effect on the photoluminescence of ZnO nanorods.

We report an investigation of the impact of surface passivation on the optical properties of ZnO nanorods. Al2O3 coating and hydrogen plasma treatment were used to passivate the surface states. It was found that Al2O3 coating led to the suppression of the deep level emissions, while hydrogen plasma treatment completely quenched the deep level emissions. It was confirmed that the surface states of the as-grown ZnO nanorod arrays indeed contributed to the deep level emissions. Evidence was also provided that shows surface states have a greater impact on the green emission than the orange emission and may cause the negative thermal quenching behavior. Moreover, the passivation effect was confirmed by the changes of the O 1s and Zn 2p spectra.

Hirsutanol A, a novel sesquiterpene compound from fungus Chondrostereum sp., induces apoptosis and inhibits tumor growth through mitochondrial-independent ROS production: hirsutanol A inhibits tumor growth through ROS production.

BACKGROUND: Hirsutanol A is a novel sesquiterpene compound purified from fungus Chondrostereum sp. in Sarcophyton tortuosum. Our previous studies had demonstrated that hirsutanol A exhibited potent cytotoxic effect on many kinds of cancer cell lines. In the current study, the antitumor activity of hirsutanol A and its molecular mechanisms were investigated. METHODS: Hirsutanol A induced growth inhibition and apoptotic cell death of human colon cancer SW620 cells and human breast cancer MDA-MB-231cells were determined using MTT assay and flow cytometry assay, respectively. The effect of hirsutanol A on intrinsic ROS level and change in mitochondrial membrane potential (△ψm) of different cell lines were also measured by flow cytometry assay. The function of JNK was compromised by JNK siRNA or JNK inhibitor SP600125. The expression of cytochrome c, p-JNK, p-c-Jun after treatment with hirsutanol A were detected by Western blot analysis. Finally, the in vivo anti-tumor effect of hirsutanol A was examined in human cancer cell SW620 xenograft model. RESULTS: The results showed that hirsutanol A significantly induced apoptosis, mitochondrial-independent increase of Reactive Oxygen Species (ROS) level, change of mitochondrial membrane potential, release of cytochrome c in human cancer cells. Preventing increase of ROS level using the potent antioxidant N-acetyl-L-cysteine (NAC) markedly decreased hirsutanol A-induced apoptosis. In addition, JNK signaling pathway was activated by hirsutanol A through elevating ROS level. Blockade of JNK signaling pathway by JNK specific inhibitor SP600125 enhanced apoptosis and hirsutanol A-induced ROS accumulation. Also, hirsutanol A exhibited antitumor activity in human cancer cell SW620 xenograft model. CONCLUSION: These data suggested that hirsutanol A inhibited tumor growth through triggering ROS production and apoptosis.

Hetero-metal cation control of CuO nanostructures and their high catalytic performance for CO oxidation.

A controllable synthesis of various morphologies of CuO nanostructures with tuning by hetero-metal cations has been developed in aqueous solution at room temperature. The morphologies of CuO can be engineered from nanosheets to nanoparticles with different length ratios of the long axis to the short axis. The formation of many metal-ion complexes plays an important role in slowing the release rate of OH(-) and affects the reaction kinetics further. We found that the effect of hetero-metal cations on the final morphology of the CuO nanostructures was the same as that of the cooling temperature. A series of temperature-controlled experiments demonstrated this. Furthermore, among all the synthesized CuO nanostructures, the fascinating colloidal mesoporous CuO quasi-monocrystalline nanosheets prepared at 25 °C with a thickness of ca. 10 nm and large specific surface area of 80.32 m(2) g(-1) is investigated intensively. These CuO nanosheets demonstrate a superior catalytic activity for CO oxidation, with features of high CO conversion efficiency (47.77 mmol(CO) g(-1)(CuO) h(-1) at 200 °C), which is close to that reported for previously investigated supported-CuO catalysts, and a low apparent activation energy E(a) (53.3 kJ mol(-1)).

Hirsutanol A induces apoptosis and autophagy via reactive oxygen species accumulation in breast cancer MCF-7 cells.

Hirsutanol A is a novel sesquiterpene compound purified from the marine fungus Chondrostereum sp in the coral Sarcophyton tortuosum. Our previous studies had demonstrated that hirsutanol A exerted potent cytotoxic effect in many kinds of cancer cell lines. Here, the anticancer molecular mechanisms of hirsutanol A were investigated in breast cancer MCF-7 cells. The results showed that hirsutanol A could inhibit cell proliferation, elevate reactive oxygen species (ROS) level, and induce apoptosis and autophagy. Co-treatment with the potent antioxidant agent N-acetyl-L-cysteine could effectively reverse the effect of enhanced ROS production, which in turn, reduces growth inhibition, apoptosis, and autophagy mediated by hirsutanol A. In addition, blocking autophagy by bafilomycin A1 or Atg7-siRNA could synergistically enhance the antiproliferative effect and apoptosis induced by hirsutanol A. These data suggested that hirsutanol A could induce apoptosis and autophagy via accumulation of ROS and co-treatment with an autophagy inhibitor could sensitize MCF-7 cells to hirsutanol A.

Dominant free exciton emission in ZnO nanorods.

Nanostructured ZnO is considered to be a promising building block in the design of nanoscale optoelectronic devices. It usually shows dominant donor-bound exciton (DX) emission at low temperatures. In this study, ZnO nanorods with high crystallinity and optical quality were grown by metal-organic chemical vapor deposition on a-plane sapphire (1120) substrates. Dominant free exciton (FX) emission at a low temperature (14 K) was observed by photoluminescence spectroscopy. It was attributed to both the enhancement of the FX emission induced by the high crystalline quality of the nanorods and the suppression of the DX emission induced by hydrogen out-diffusion. The latter reason is believed to be more important from the analysis of the hydrogen distribution in the nanorods through photoluminescence spectroscopy and secondary ion mass spectrometry. A slow cooling process during the deposition is suggested to result in a better optical quality. These results can promote our understanding of the optical properties of ZnO nanostructures.

A novel sesquiterpene Hirsutanol A induces autophagical cell death in human hepatocellular carcinoma cells by increasing reactive oxygen species.

BACKGROUND AND OBJECTIVE: Hirsutanol A is a novel sesquiterpene compound purified from fungus chondrostereum sp in Sarcophyton tortuosum. Its pharmacologic effect has not been reported yet. This study aimed to investigate cytotoxic effect of Hirsutanol A on hepatocellular carcinoma (HCC) cells and its mechanism. METHODS: Hep3B cells were treated with different concentrations of Hirsutanol A. Cell proliferation was detected by MTT assay. The protein expression of LC3 was determined by Western blot. The generation of reactive oxygen species (ROS) was monitored by flow cytometry. RESULTS: Hirsutanol A significantly inhibited proliferation of Hep3B cells with 50% inhibition concentrations (IC50) of 14.54, 6.71, and 3.59 micromol/L when exposed to Hirsutanol A for 24, 48, and 72 h, respectively. Incubation of Hep3B cells with Hirsutanol A markedly increased the level of ROS and the autophagy marker MAP-LC3 conversion from type I to type II. Pre-incubation with an antioxidant N-acetyl cystein (NAC) decreased the level of ROS, and reduced MAP-LC3 I-II conversion, and suppressed cell death. Blocking autophagy with a specific autophagy inhibitor 3-methyladenine (3-MA), the cytotoxic effect of this compound was attenuated. CONCLUSION: Hirsutanol A has potent cytotoxic effect, and can induce autophagic cell death via increasing ROS production.