Search

This paper studies audio-visual noise suppression for egocentric videos -- where the speaker is not captured in the video. Instead, potential noise sources are visible on screen with the camera emulating the off-screen speaker's view of the outside world. This setting is different from prior work in audio-visual speech enhancement that relies on lip and facial visuals. In this paper, we first demonstrate that egocentric visual information is helpful for noise suppression. We compare object recognition and action classification-based visual feature extractors and investigate methods to align audio and visual representations. Then, we examine different fusion strategies for the aligned features, and locations within the noise suppression model to incorporate visual information. Experiments demonstrate that visual features are most helpful when used to generate additive correction masks. Finally, in order to ensure that the visual features are discriminative with respect to different noise types, we introduce a multi-task learning framework that jointly optimizes audio-visual noise suppression and video-based acoustic event detection. This proposed multi-task framework outperforms the audio-only baseline on all metrics, including a 0.16 PESQ improvement. Extensive ablations reveal the improved performance of the proposed model with multiple active distractors, overall noise types, and across different SNRs., Accepted to ICASSP 2023

Although ovarian cancer, a gynecological malignancy, has the highest fatality rate, it still lacks highly specific biomarkers, and the differential diagnosis of ovarian masses remains difficult to determine for gynecologists. Our study aimed to obtain ovarian cancer-specific protein candidates from the circulating small extracellular vesicles (sEVs) and develop a protein panel for ovarian cancer screening and differential diagnosis of ovarian masses. In our study, sEVs derived from the serum of healthy controls and patients with cystadenoma and ovarian cancer were investigated to obtain a cancer-specific proteomic profile. In a discovery cohort, 1119 proteins were identified, and significant differences in the protein profiles of EVs were observed among groups. Then, 23 differentially expressed proteins were assessed using the parallel reaction monitoring in a validation cohort. Through univariate and multivariate logistic regression analyses, a novel model comprising three proteins (fibrinogen gamma gene (FGG), mucin 16 (MUC16), and apolipoprotein (APOA4)) was established to screen patients with ovarian cancer. This model exhibited an area under the receiver operating characteristic curve (AUC) of 0.936 (95% CI, 0.888–0.984) with 92.0% sensitivity and 82.9% specificity. Another panel comprising serum CA125, sEV-APOA4, and sEV-CD5L showed excellent performance (AUC 0.945 (95% CI, 0.890–1.000), sensitivity of 88.0%, specificity of 93.3%, and accuracy of 89.2%) to distinguish malignancy from benign ovarian masses. Altogether, our study provided a proteomic signature of circulating sEVs in ovarian cancer. The diagnostic proteomic panel may complement current clinical diagnostic measures for screening ovarian cancer in the general population and the differential diagnosis of ovarian masses.

Objective To develop a novel diagnostic nomogram model to predict malignancy in patients with ovarian masses. Methods In total, 1277 patients with ovarian masses were retrospectively analyzed. Receiver operating characteristic (ROC) analysis was performed to identify valuable predictive factors. Univariate and multivariate logistic regression analyses were used to identify risk factors for ovarian cancer. Subsequently, a predictive nomogram model was developed. The performance of the nomogram model was assessed by its calibration and discrimination in a validation cohort. Decision curve analysis (DCA) was applied to assess the clinical net benefit of the model. Results Overall, 496 patients (38.8%) had ovarian cancer. Eighteen parameters were significantly different between the malignant and benign groups. Five parameters were identified as being most optimal for predicting malignancy, including age, carbohydrate antigen 125, fibrinogen-to-albumin ratio, monocyte-to-lymphocyte ratio, and ultrasound result. These parameters were incorporated to establish a nomogram model, and this model exhibited an area under the ROC curve (AUC) of 0.937 (95% confidence interval [CI], 0.920–0.954). The model was also well calibrated in the validation cohort and showed an AUC of 0.925 (95%CI, 0.896–0.953) at the cut-off point of 0.298. DCA confirmed that the nomogram model achieved the best clinical utility with almost the entire range of threshold probabilities. The model has demonstrated superior efficacy in predicting malignancy compared to currently available models, including the risk of ovarian malignancy algorithm, copenhagen index, and the risk of malignancy index. More importantly, the nomogram established here showed potential value in identification of early-stage ovarian cancer. Conclusion The cost-effective and easily accessible nomogram model exhibited favorable accuracy for preoperative prediction of malignancy in patients with ovarian masses, even at early stages.

Deep neural networks have been successfully applied to sound direction-of-arrival estimation under challenging conditions. However, such a learning-based approach requires a large amount of labeled training data, which is difficult to acquire. To address this problem, we propose a novel approach for multi-speaker direction-of-arrival estimation with data augmentation and weakly-supervised domain adaptation. We generate source domain data with simulation, and collect real data annotated with the number of sound sources as the weak labels. The real data are further augmented by mixing single-source segments. Then, weakly-supervised domain adaptation is applied to models pre-trained on the simulated data. We define a loss function for the adaptation process which exploits the weak labels and the mixture component information in the augmented data. Experiments with real robot audio data show that our proposed approach achieves similar performance as if the fully-labeled real data are used. This paper suggests an effective development procedure for DOA estimation models applied to new types of microphone arrays with minimal data collection efforts.

Most microplastic particles may undergo various aging in water environments. In this work, surface physicochemical properties were firstly compared among pristine polypropylene (PP-pris) microplastics, and two aged ones obtained after pretreated with HCl (PP-acid) and NaOH (PP-alka). When compared with PP-pris and PP-acid, PP-alka had a much stronger Mn(II) adsorption capacity. The results regarding the role of natural organic matter and colloidal particle concentrations on adsorption demonstrated that for water solutions either containing kaolin or not, humic acid (HA) had significantly negative influence on Mn(II) adsorption capacity of PP-alka due to their complexation and competition effects, and its negative influence became enhanced with increasing kaolin concentrations. Besides, established conceptual models of adsorption were applied to comprehensively explore adsorption mechanisms of PP-alka for Mn(II) in the coexistence of HA and kaolin. An important suggestion was that in complicated adsorption-reactor system, great numbers of microplastics-kaolin heteroaggregates might be formed via ion bridging of Mn(II) and/or polymer bridging of HA. So these formed aggregates were possible to re-organize themselves, under pre-set vibration-speed conditions, for achieving a more stable structure. As a consequence, Mn(II) adsorption behaviors would be affected by changes in steric-hindrance effects of HA molecules and surface charge distribution of resultant heteroaggregates.

The 5-year survival rate of ovarian cancer patients is only 47%, and developing novel drugs for ovarian cancer is needed. Herein, we evaluated if and how SRT2183, a sirtuin-1 activator, impairs the ovarian cancer cells. OVCAR-3 and A2780 cells were treated with SRT2183. Cell viability was measured by cell counting kit-8 assay and clonogenic assay. Apoptosis was determined by flow cytometry with Annexin V and propidium iodide. The level of autophagy was evaluated by western blot and immunofluorescence. The activities of AKT/mTOR/70s6k and MAPK signaling pathway were measured by immunoblot. SRT2183 inhibited the growth of ovarian cancer cells, increased the accumulation of BAX, cleaved-caspase 3 and cleaved-PARP, and decreased the level of anti-apoptotic Bcl-2 and Mcl-1. SRT2183 increased the LC3II level, and enhanced the degradation of p62/SQSTM1. SRT2183 increased the formation of GFP-LC3 puncta and induced the maturation of autophagosome. Interestingly, knockdown of autophagy related 5 and 7 significantly impaired the anti-carcinoma activity of SRT2183, implying that SRT2183 impaired the ovarian cancer cells by inducing autophagy. SRT2183 decreased the accumulation of p-Akt, p-mTOR and p-70s6k, and activated the p38 MAPK signaling pathway. This indicated that Akt/mTOR/70s6k and p38 MAPK signaling pathway might be involved in the SRT2183-mediated autophagy and apoptosis.

Chromodomain helicase DNA binding protein 1-like (CHD1L) gene has been proposed to play an oncogenic role in human hepatocellular carcinoma. Previously we reported that CHD1L overexpression is significantly associated with the metastasis proceeding of epithelial ovarian cancer (EOC), and may predict a poor prognosis in EOC patients. However, the potential oncogenic mechanisms by which CHD1L acts in EOC remain unclear. To elucidate the oncogenic function of CHD1L, we carried out a series of in vitro assays, with effects of CHD1L ectogenic overexpression and silencing being determined in EOC cell lines (HO8910, A2780 and ES2). Real-time PCR and Western blotting analyses were used to identify potential downstream targets of CHD1L in the process of EOC invasion and metastasis. In ovarian carcinoma HO8910 cell lines, ectopic overexpression of CHD1L substantially induced the invasive and metastasis ability of the cancer cells in vitro. In contrast, knockdown of CHD1L using shRNA inhibited cell invasion in vitro in ovarian carcinoma A2780 and ES2 cell lines. We also demonstrated that methionyl aminopeptidase 2 (METAP2) was a downstream target of CHD1L in EOC, and we found a significant, positive correlation between the expression of CHD1L and METAP2 in EOC tissues (P

Within the stirred-tank flocculating system under equivalent shear-rate (Gave) cases, the effects of both reactor shape (cylindrical or square) and baffles on floc morphological evolution during flocculation were investigated by applying flocculation tests with the aid of computational fluid dynamics (CFD) simulations. At a constant Gave, the recirculating times for those floc aggregates existing in water flow to the impeller region and the intensity of turbulence occurring within this region, together with the non-uniform nature of turbulence for water flow above the rotating impeller, were found to be distinct in the various stirred-tank reactors investigated, thereby affecting the development of floc size and structure. Besides, the importance of reactor shape and baffles to floc morphological evolution seemed to be largely determined by predominant growth behavior(s). For aggregation-dominated circumstances, the baffled square stirred tank formed the largest and most compact floc aggregates among all reactors, due to the most visiting times for pre-formed micro-floc aggregates to the highest turbulent-intensity impeller region; while for breakage-dominated circumstances, the most non-uniform distributions of turbulent water flow were generated via impeller-baffle interactions in this reactor, thereby giving the highest breakage rate and producing the smallest floc aggregates with the smoothest and most compact structures.

The impeller clearance (C) and shear rate (Gave) effects on aggregation and breakage behaviors in the whole process of floc growth were investigated by applying lab-scale flocculation tests with the aid of computational fluid dynamics (CFD) simulations. For a baffled square stirred-tank flocculating reactor, the recirculating times for water flow to the impeller region and intensity of turbulence occurring within this region, together with the non-uniform nature of turbulence for water flow above the rotating impeller, were found to be more or less different in all C-value cases at a constant Gave, thereby affecting the development of floc morphology. Besides, the importance of impeller clearance (C) to floc morphological evolution seemed to be largely determined by predominant growth behavior(s), likely as a result of distinct occurring degrees of either aggregation or breakage for each combination of C and Gave; and interestingly, a certain compaction for the above effect took place at a relatively high shear, possibly attributed to increasingly pronounced occurrence of floc breakage and irreversibility of this behavior. To produce fractal flocs with favorable morphological characteristics, the impeller should be placed at an appropriate distance from the reactor bottom within the flocculating system, under whether aggregation-dominated or breakage-dominated circumstances, for weakening the role of reactor bottom and water surface on floc growth behaviors during flocculation. This research would be beneficial for thoroughly understanding floc formation mechanisms and optimally designing impeller-based flocculating reactors.

For enhancing the understanding of ballasted flocculation (BF) mechanisms, a series of BF tests were performed to explore the coagulant dosage and mixing intensity effects on floc growth behaviors and corresponding particle settling by using coagulants of aluminum sulphate (AS) or polyaluminum chloride (PACl). As expected, different coagulant dosages brought about distinct surface charge characteristics of micro-floc aggregates after either Al-based coagulant addition, thereby affecting the development of ballasted floc size. More specifically, the higher the coagulant dosage was, the more the cationic hydrolyzed-Al species were produced, leading to a requirement of more binding sites on the injected polymer for bridging between micro-flocs and ballasting agent (BA) particles, and accordingly, the degree to which BA particles had been incorporated into the aggregate matrix seemed to largely account for ballasted floc formation. Besides, floc aggregates formed after breakage appeared larger than those after maturation, especially at higher coagulant dosages, possibly caused by a stretched conformation of long polymer chains providing more binding sites available within a higher-shear suspension. In an operational sense, poor incorporation should be maximally eliminated, since this undesirable behavior could produce a greater number of resultant flocs with fewer or even no BA particles, thus worsening the settling performance. This research would be helpful for thoroughly understanding ballasted floc formation mechanisms in either water or wastewater treatment operations.

The IEEE Spoken Language Technology Workshop (SLT) 2021 Alpha-mini Speech Challenge (ASC) is intended to improve research on keyword spotting (KWS) and sound source location (SSL) on humanoid robots. Many publications report significant improvements in deep learning based KWS and SSL on open source datasets in recent years. For deep learning model training, it is necessary to expand the data coverage to improve the robustness of model. Thus, simulating multi-channel noisy and reverberant data from single-channel speech, noise, echo and room impulsive response (RIR) is widely adopted. However, this approach may generate mismatch between simulated data and recorded data in real application scenarios, especially echo data. In this challenge, we open source a sizable speech, keyword, echo and noise corpus for promoting data-driven methods, particularly deep-learning approaches on KWS and SSL. We also choose Alpha-mini, a humanoid robot produced by UBTECH equipped with a built-in four-microphone array on its head, to record development and evaluation sets under the actual Alpha-mini robot application scenario, including noise as well as echo and mechanical noise generated by the robot itself for model evaluation. Furthermore, we illustrate the rules, evaluation methods and baselines for researchers to quickly assess their achievements and optimize their models., Accepted at IEEE SLT 2021

The prognosis for cervical cancer (CCa) patients with lymph node metastasis (LNM) is dismal. Elucidation of the molecular mechanisms underlying LNM may provide clinical therapeutic strategies for CCa patients with LNM. However, the precise mechanism of LNM in CCa remains unclear. Herein, we demonstrated that protein tyrosine phosphatase receptor type M (PTPRM), identified from TCGA dataset, was markedly upregulated in CCa with LNM and correlated with LNM. Moreover, PTPRM was an independent prognostic factor of CCa patients in multivariate Cox′s proportional hazards model analysis and associated with poor prognosis. Furthermore, through gain-of-function and loss-of-function approaches, we found that PTPRM promoted CCa cells proliferation, migration, invasion, lymphangiogenesis, and LNM. Mechanistically, PTPRM promoted epithelial–mesenchymal transition (EMT) via Src-AKT signaling pathway and induced lymphangiogenesis in a VEGF-C dependent manner, resulting in LNM of CCa. Importantly, knockdown of PTPRM dramatically reduced LNM in vivo, suggesting that PTPRM plays an important role in the LNM of CCa. Taken together, our findings uncover a novel molecular mechanism in the LNM of CCa and identify PTPRM as a novel prognostic factor and potential therapeutic target for LNM in CCa.

Background Genetically engineered mice are ideal models to advance our understanding the tumorigenesis of ovarian cancer. Our original objective was to establish an ovarian cancer model induced by Kras activation and Pten deletion. However, proficiently establishing the model remains a technical problem, which limits its application. Methods We established the Kras activation/Pten deletion-induced mouse model of ovarian cancer by injecting Cre recombinase-expressing adenovirus in the ovarian bursa. PCR analysis, Western blotting, and immunohistochemistry staining were performed to verify the alteration of conditional genes. We detected expression of canonical molecular markers in order to examine the origin of the tumors. Results Subcutaneous lumps developed accidentally in mice with ovarian cancer, as early as 2 weeks post in vivo genetic manipulation, far before the destructive growth of ovarian cancer. PCR analysis confirmed the efficient Cre-mediated recombination of Kras and Pten in tumor tissues, which are consistent with the activation of the MAPK and PI3K/Akt/mTOR pathways. Histomorphological and histological analysis showed that the lumps were actually rhabdomyosarcoma (RMS). We confirmed that the leakage of adenovirus transformed healthy adjacent tissues into RMS. Conclusions Avoiding accidental exposure of non-target tissues to adenovirus is crucial to successfully establish the ovarian cancer mouse model. Moreover, non-specific genetic manipulations can induce the development of RMS.

This study focuses on the effect of mixing hydrodynamics, generated in unbaffled square stirred tanks with different liquid-level heights (H), on floc growth during flocculation over the same shear-rate range (Gave = 10–70 s−1) by using an in-situ recognition system for floc morphology. The hydrodynamic environments under each combination of H and Gave were characterized by calculated circulation time (tc) and impeller tip speed (Nd). It was found that the temporal evolution of floc size appeared to be sensitive to the value of H at a given Gave, and with increasing Gave, greater difference occurred at the same flocculation moments for different H-value cases. Also, under a fixed Gave-value condition, the H-value effect on floc size evolution at steady state of flocculation was shown to largely depend upon the predominant growth mechanism(s). An appropriate liquid-level height should be required to bring about an optimal combination of tc and Nd, thereby maximizing floc growth rate under both aggregation-dominated and breakage-dominated circumstances for enhancing floc formation and subsequent particle removal efficiency. The present study may provide useful insights into the design and operation of stirred-tank reactors for flocculation.

The effect of liquid-level height (H) on floc growth during flocculation has been investigated in unbaffled square stirred tanks with a fixed bottom width (D). Firstly, flocculation tests were performed by using an in situ recognition system for floc morphology to evaluate flocculation performance within each tank at three typical shear rates of Gave = 10, 30 and 70 s−1. Then, turbulent flow fields generated under all flocculation-test conditions were predicted by computational fluid dynamics (CFD) simulations, followed by a detailed discussion based on experimental and numerical data. It was found that stirred tanks with different H values (characterized by H-to-D ratio) caused distinct turbulent flow fields at the same shear rates examined, reflected by different turbulent intensities in the impeller zone and different non-uniform nature of turbulence in the overall flocculating system. Accordingly, the evolution of floc size and structure was greatly affected due to various aggregation and breakage kinetics. Moreover, the effect of H-to-D ratio on floc growth appeared to be related to the predominant growth mechanism(s), i.e., for the shear rates of Gave = 10 and 30 s−1, little or no significant breakage occurred and more significant non-uniformity of turbulence generated in the stirred tank with a higher H-to-D ratio led to increasingly pronounced behavior of restructuring, forming aggregates with more compact structure caused by primary-particle rearrangement, whereas at Gave = 70 s−1, breakage dominated over aggregation and resultant aggregates became smaller and more compact with increasing H-to-D ratios, possibly due to the irreversibility of floc breakage. Therefore, an appropriate liquid-level height of tank should be designed to enhance floc formation and subsequent particle removal efficiency. The combined means of flocculation tests and CFD simulations may provide useful insights for optimizing the design and operation of stirred-tank reactors for flocculation.

This work mainly focuses on the effect of baffle width ( B ) on floc growth within square stirred-tank reactors for flocculation. Firstly, a series of flocculation tests were carried out, using polyaluminum chloride (PACl) as coagulant, to evaluate flocculation performance in each tank at three typical shear rates of G ave = 10, 30 and 70 s −1 . Experimental results were reported in terms of kaolin-floc average size, size distributions and perimeter-based fractal dimension, derived from an in-situ recognition system for floc morphology. Then, the hydrodynamic environments (where the growth of floc took place) were characterized, by performing Computational Fluid Dynamics (CFD) simulations, with predicted results of area-weighted average turbulent kinetic energy and its dissipation rate in the plane containing the impeller, and circulation time (denoted as k 0.33 H , e 0.33 H and t c , respectively), followed by a detailed discussion based on experimental and numerical results. As expected, baffles with different widths caused distinct turbulent flow fields in corresponding stirred tanks at the same shear rates, thereby affecting the evolution of floc size and structure during flocculation. According to CFD predictions, for a constant G ave , the baffle width of B = 0.10 D (where D is the bottom width of tank) gave rise to the highest values of k 0.33 H and e 0.33 H , as well as the shortest t c , followed by B = 0.13 D , 0 (unbaffled) and 0.07 D , and for B = 0.20 D , the lowest k 0.33 H and e 0.33 H , together with the longest t c , were produced. It was found that the floc growth dependence on baffle width appeared to be greatly related to the predominant growth mechanism(s), i.e., at the shear rates of G ave = 10 and 30 s −1 , where the rate of breakage was approximately zero (because little or unimportant breakage was observed), a higher e 0.33 H and a short t c corresponded to an increased rate of floc growth resulting from higher particle collision frequency, thus forming larger flocs under the same shear-rate conditions, whereas at G ave = 70 s −1 , where breakage dominated over aggregation, larger flocs seemed to be produced by a lower e 0.33 H and a longer t c , likely as a result of a lower breakage rate and more sufficient opportunity for broken aggregates to reform or restructure during water flow circulation. Moreover, the effect on floc size and structure during flocculation was somewhat compressed when breakage became pronounced, and the same was true of the restructuring behavior. To enhance floc growth process, some baffles should be installed to break water flow co-rotation with the impeller (to increase axial flow rate), and however, larger-area dead zones might be simultaneously formed. Therefore, an appropriate width (e.g., B = 0.10 D or 0.13 D ) would be required to maintain a compromise. The present study may provide useful insights for optimizing the design and operation of baffled stirred-tank reactors for flocculation.

In this paper, we introduce spatial attention for refining the information in multi-direction neural beamformer for far-field automatic speech recognition. Previous approaches of neural beamformers with multiple look directions, such as the factored complex linear projection, have shown promising results. However, the features extracted by such methods contain redundant information, as only the direction of the target speech is relevant. We propose using a spatial attention subnet to weigh the features from different directions, so that the subsequent acoustic model could focus on the most relevant features for the speech recognition. Our experimental results show that spatial attention achieves up to 9% relative word error rate improvement over methods without the attention., Comment: To be presented at ICASSP 2020

Despite the recent success of deep neural network-based approaches in sound source localization, these approaches suffer the limitations that the required annotation process is costly, and the mismatch between the training and test conditions undermines the performance. This paper addresses the question of how models trained with simulation can be exploited for multiple sound source localization in real scenarios by domain adaptation. In particular, two domain adaptation methods are investigated: weak supervision and domain-adversarial training. Our experiments show that the weak supervision with the knowledge of the number of sources can significantly improve the performance of an unadapted model. However, the domain-adversarial training does not yield significant improvement for this particular problem.

This paper presents a universal partially evolved parallelization of the multi-objective evolutionary algorithm based on decomposition (MOEA/D) for multi-objective optimization on message-passing clusters to reduce its computation time. The partially evolved MOEA/D (peMOEA/D) is suitable not only for the bi-objective space, but also for higher dimensional objective spaces by using a partially evolved island model. This model improves the algorithm universality and population diversity by keeping a subpopulation equal in size to the entire population, but only evolving a partial (equal to a partition size) subpopulation on each separate processor in a cluster. Furthermore, a nearest-neighbor partitioning approach, hybrid migration policy and adaptive neighbor-based topology are adopted in the peMOEA/D. The fat partitions generated by the nearest-neighbor partitioning can reduce migration traffic of elitist individuals across subpopulations. Then, hybrid migration of both elitist individuals and utopian points helps separate subpopulations to cooperate in guiding the search quickly towards a complete front. Next, the adaptive neighbor-based topology connects neighbor partitions only and achieves an excellent balance between convergence speed and migration traffic. Experimental results on benchmark multi-objective optimization problems with two or more objectives demonstrate the satisfactory overall performance of the peMOEA/D in terms of both convergence performance and speedup on message-passing clusters.

Sb(V) can cause chronic or acute poisoning to animals and plants; Unfortunately, traditional coagulation–flocculation–sedimentation (CFS) process has limited effect on Sb removal. In this study, flocs produced by CF process were preloaded on the surface of the UF membrane, and the coordinated coagulation-floc preloaded-ultrafiltration (CFPLU) process was proposed. The removal effect of the CFPLU process on composite pollutants and the fouling degree of the UF membrane were studied under different solid contents of floc-preloaded water samples. Moreover, the removal mechanism of Sb(V) and membrane fouling mechanism in CFPLU process were also investigated. The results indicated that, compared with the CFS process, UF process, CF ultrafiltration (CF-UF) process, preloaded UF (PL-UF) process, the CFPLU process has the highest removal rate for Sb(V) and the lowest degree of membrane fouling. When the solid content is 1.5%, Sb(V) removal rate reaches the highest value of 97.29% and Sb(V) concentrations of CFPLU under all solid contents are

This paper presents a 6-18 GHz active 6-bit phase shifter based on a vector-sum technique in $0.13-\mu m$ SiGe BiCMOS technology for X- and Ku-band phased arrays. An input Marchand balun and an L-C resonance-based quadrature all-pass filter are used to generate two orthogonal vectors with high $I/Q$ accuracy and high linearity over a wide frequency band. The $I/Q$ reference signals are then fed into two VGAs, of which the gains are adjusted with an integrated current-mode DAC to achieve the 6-bit phase resolution between $0-360^{o}$. The phase shifter shows 3.12 $dB$ of power gain at 9.25 GHz with a 3-dB gain bandwidth of 6.4-12.9 GHz, and the RMS gain variation is $\lt 1.05 dB$ for all 6-bit phase states. The phase shifter achieves an input P1dB of $5-8dBm$ at 6-18 GHz for 0°-phase state. The measured RMS phase error is $\lt 5.6^{o}$ at 5-18 GHz. The total current consumption is 28.2 $mA$ from a 3.3 V supply voltage and the overall chip size is $1.85\times 1.32mm^{2}.$

This paper presents a wideband power amplifier in 0.13-µm SiGe BiCMOS technology for 26.5–40GHz Ka-band application. A 6 order input matching network and stacked cascode structure with interstage neutral inductor are used for bandwidth expansion. The simulation results show the gain is larger than 16.1dB, return loss is less than −10dB and P1dB output powers is larger than 19.8dBm in the whole Ka band. The area of the power amplifier is 1.2×1.05 mm2, with a total current consumption of 178 mA from a 6 V supply voltage.

An Ultra-Wideband low-noise amplifier LNA designed in 0.13μm SiGe BiCMOS technology is presented in this paper. The inductive peaking technology is adopted to expand the bandwidth, the resistance negative feedback technology and the emitter degeneration inductive technology are adopted to improve the flatness of gain and input matching. This LNA achieves a flat gain of 14.7~16.3dB in the frequency range of 6~18GHz. The noise figure (NF) of the LNA is 2.7~3.8 dB across the band. The input return losses (S11) of the LNA are better than −16.4dB and the output return losses (S22) of the LNA are better than −14.9dB. The input 1dB compression point (IP1dB) of the LNA is −5.0dBm. The LNA dissipates 9.8mA with a 3.3V supply.

We propose a novel multi-task neural network-based approach for joint sound source localization and speech/non-speech classification in noisy environments. The network takes raw short time Fourier transform as input and outputs the likelihood values for the two tasks, which are used for the simultaneous detection, localization and classification of an unknown number of overlapping sound sources, Tested with real recorded data, our method achieves significantly better performance in terms of speech/non-speech classification and localization of speech sources, compared to method that performs localization and classification separately. In addition, we demonstrate that incorporating the temporal context can further improve the performance.

Epithelial ovarian cancer (EOC) is the most lethal gynecologic malignancy with high recurrence and poor prognosis duo to the lack of effective biomarkers. TBC1 domain family member 16 (TBC1D16), a GTPase-activating protein, is involved in regulating intracellular trafficking in tumorigenesis and metastasis. However, the clinical significance of TBC1D16 in EOC remains unknown. In the present study, we investigated the expression and prognostic significance of TBC1D16 in EOC and its relationship with the expression of vascular endothelial growth factor (VEGF). The tissue specimens included 156 histologically confirmed EOC and 30 normal ovarian tissues. The expression of TBC1D16 and VEGF was detected by immunohistochemistry (IHC), and the immunoreactive score was calculated with signal intensity and percentage of positive cells. IHC results showed that TBC1D16 and VEGF were both mainly localized in cytoplasm of epithelial cells in normal ovarian tissues and were expressed in cancer cells. Based on the immunoreactive score, TBC1D16 expression in EOC was categorized as "high expression," compared with normal ovarian tissues (P < 0.05). The Chi-square test showed that high TBC1D16 expression was related to advanced pT stages (P = 0.029), but not correlated with other clinical features. Moreover, the TBC1D16 expression was significantly higher in EOC specimens with low VEGF expression (P < 0.001). Importantly, in both univariate and multivariate survival analyses, high expression of TBC1D16 was significantly correlated with good overall survival (OS). In conclusion, TBC1D16 is a predictive marker for favorable prognosis of EOC.

The idea of decomposition is becoming increasingly successful and popular in evolutionary multi-objective optimization. An efficient cone decomposition approach was further developed in the conical area evolutionary algorithm (CAEA). This approach improves the runtime efficiency and population diversity of decomposition-based algorithms effectively for bi-objective optimization in practice. In this paper, it is proved that the optimum of any conical subproblem in cone decomposition possesses the global Pareto optimality, apart from previously known the local Pareto optimality, in the presence of continuous frontier segment within the associated subregion. Experimental results on 5 bi-objective benchmark problems indicate that the solutions of conical subproblems obtained by CAEA not only acquire obviously better Pareto optimality in terms of C-metric but also constitute higher qualities of frontiers than the solutions by the other four popular multi-objective algorithms.

Clusterin (CLU), a multifunctional glycoprotein, is ubiquitously produced in mammalian tissues. CLU has been shown to play significant roles in many of the biological behaviours of human tumors, such as cell proliferation, apoptosis, chemoresistance and angiogenesis. However, the relationship of CLU expression with angiogenesis in ovarian cancer has not been studied. A total of 275 epithelial ovarian tumors were obtained from archives of paraffin‑embedded tissues. Immunohistochemical (IHC) staining for CLU and vascular endothelial growth factor (VEGF) was performed on a tissue microarray (TMA) including 181 primary ovarian epithelial cancer, 40 borderline ovarian tumors and 54 ovarian cancer mesenteric metastasis samples. Of the 174 cases, overexpression of CLU and VEGF were detected in 107 (61.5%) and 109 (62.9%) cases of primary ovarian carcinoma, respectively. Of the 107 cases of primary ovarian carcinoma with overexpression of CLU, expression of VEGF was increased in 82 (75.2%) cases. However, in another 67 cases without CLU overexpression, overexpression of VEGF was observed in only 27 (24.8%) cases (P

To characterize abnormal epigenetic changes and protein expression of the clusterin gene in a large series of ovarian malignant and borderline tumors. Protein expression and promoter methylation of clusterin gene in 181 primary ovarian epithelial cancer, 40 borderline ovarian tumors, 54 ovarian cancer mesenteric metastasis, and 10 normal ovarian samples were analyzed by immunohistochemical staining and methylation-specific PCR. Overexpression of clusterin protein was frequently seen in various ovarian epithelial tumors, being detected in 102 of 181 (56 %) primary ovarian epithelial cancers, 21 of 37 (57 %) borderline ovarian tumors. Surprisingly, clusterin protein expression was significantly reduced in mesenteric metastasis (20 of 54; 37 % cases), as compared to primary ovarian carcinoma (p = 0.01). Overexpression of clusterin protein was significantly correlated with high-grade histology (p = 0.002) and high FIGO stages (p = 0.05). Clusterin promoter hypermethylation was detected in 24 of 181 (13 %) primary ovarian epithelial cancer, 8 of 54 (14 %) mesenteric metastasis, and 10 of 37 (27 %) borderline ovarian tumors. Overall, clusterin promoter hypermethylation was significantly correlated with decreased protein expression in these samples (p