Search

In recent years, the fast rise in number of studies on graph neural network (GNN) has put it from the theories research to the real-world application stage. Despite the encouraging performance achieved by GNN, less attention has been paid to the privacy-preserving training and inference over distributed graph data in the related literature. Due to the particularity of graph structure, it is challenging to extend the existing private learning frameworks to GNN. Motivated by the idea of split learning, we propose a server aided privacy-preserving GNN (SAPGNN) for the intra-graph node level task on the horizontally partitioned cross-silo scenario. It offers a natural extension of centralized GNN to the isolated graph with max/min pooling aggregation, while guaranteeing that all the private data involved in the computation still stays with local data holders. To further enhance the data privacy, a secure pooling aggregation mechanism is proposed. Theoretical and experimental results show that the proposed model achieves the same accuracy as the one learned over the combined data.

Abstract Loss of ARID1A, a subunit of the SWI/SNF chromatin remodeling complex, contributes to malignant progression in multiple cancers including non-small cell lung cancer (NSCLC). In the search for key genes mediating the aggressive phenotype caused by ARID1A loss, we analyzed 3 Gene Expression Omnibus (GEO) datasets that contain RNA sequencing data from ARID1A-depleted cancer cells. PLAU was identified as a common gene that was induced in different cancer cells upon ARID1A depletion. Overexpression of PLAU positively modulated NSCLC cell growth, colony formation, cisplatin resistance, and survival under serum deprivation. Moreover, enforced expression of PLAU enhanced tumorigenesis of NSCLC cells in nude mice. Mechanistically, PLAU interacted with TM4SF1 to promote the activation of Akt signaling. TM4SF1-overexpressing NSCLC cells resembled those with PLAU overepxression. Knockdown of TM4SF1 inhibited the growth and survival and increased cisplatin sensitivity in NSCLC cells. The interaction between PLAU and TM4SF1 led to the activation of Akt signaling that endowed ARID1A-depleted NSCLC cells with aggressive properties. In addition, treatment with anti-TM4SF1 neutralizing antibody reduced the growth, cisplatin resistance, and tumorigenesis of ARID1A-depleted NSCLC cells. Taken together, PLAU serves as a target gene of ARID1A and promotes NSCLC growth, survival, and cisplatin resistance by stabilizing TM4SF1. Targeting TM4SF1 may be a promising therapeutic strategy for ARID1A-mutated NSCLC.

Background & purpose: For the best possible outcomes from therapy, proximal femur bone cancers must be accurately classified. This work creates an artificial intelligence (AI) model based on plain radiographs to categorize bone tumor in the proximal femur. Materials and methods: A tertiary referral center's standard anteroposterior hip radiographs were employed. A dataset 538 images of the femur, including malignant, benign, and tumor-free cases, was employed for training the AI model. There is a total of 214 images showing bone tumor. Pre-processing techniques were applied, and DenseNet model utilized for classification. The performance of the DenseNet model was compared to that of human doctors using cross-validation, further enhanced by incorporating Grad-CAM to visually indicate tumor locations. Results: For the three-label classification job, the suggested method boasts an excellent area under the receiver operating characteristic (AUROC) of 0.953. It scored much higher (0.853) than the diagnosis accuracy of the human experts in manual classification (0.794). The AI model outperformed the mean values of the clinicians in terms of sensitivity, specificity, accuracy, and F1 scores. Conclusion: The developed DenseNet model demonstrated remarkable accuracy in classifying bone tumors in the proximal femur using plain radiographs. This technology has the potential to reduce misdiagnosis, particularly among non-specialists in musculoskeletal oncology. The utilization of advanced deep learning models provides a promising approach for improved classification and enhanced clinical decision-making in bone tumor detection.

Nonnegative matrix factorization (NMF) and its numerous variants have been extensively studied and used in hyperspectral unmixing (HU). With the aid of the designed deep structure, deep NMF-based methods demonstrate advantages in exploring the hierarchical features of complex data. However, a noise corruption problem commonly exists in hyperspectral data and severely degrades the unmixing performance of deep NMF-based methods when applied to HU. In this study, we propose an ℓ2,1 norm-based robust deep nonnegative matrix factorization (ℓ2,1-RDNMF) for HU, which incorporates an ℓ2,1 norm into the two stages of the deep structure to achieve robustness. The multiplicative updating rules of ℓ2,1-RDNMF are efficiently learned and provided. The efficiency of the presented method is verified in experiments using both synthetic and genuine data.

Abstract Background ZNF674-AS1, a recently characterized long noncoding RNA, shows prognostic significance in hepatocellular carcinoma and glioma. However, the expression and function of ZNF674-AS1 in non-small cell lung cancer (NSCLC) are unclear. Methods In this work, we investigated the expression of ZNF674-AS1 in 83 pairs of NSCLC specimens and adjacent noncancerous lung tissues. The clinical significance of ZNF674-AS1 in NSCLC was analyzed. The role of ZNF674-AS1 in NSCLC growth and cell cycle progression was explored. Results Our data show that ZNF674-AS1 expression is decreased in NSCLC compared to normal tissues. ZNF674-AS1 downregulation is significantly correlated with advanced TNM stage and decreased overall survival of NSCLC patients. Overexpression of ZNF674-AS1 inhibits NSCLC cell proliferation, colony formation, and tumorigenesis, which is accompanied by a G0/G1 cell cycle arrest. Conversely, knockdown of ZNF674-AS1 enhances the proliferation and colony formation of NSCLC cells. Biochemically, ZNF674-AS1 overexpression increases the expression of p21 through downregulation of miR-423-3p. Knockdown of p21 or overexpression of miR-423-3p blocks ZNF674-AS1-mediated growth suppression and G0/G1 cell cycle arrest. In addition, ZNF674-AS1 expression is negatively correlated with miR-423-3p in NSCLC specimens. Conclusions ZNF674-AS1 suppresses NSCLC growth by downregulating miR-423-3p and inducing p21. This work suggests the therapeutic potential of ZNF674-AS1 in the treatment of NSCLC.

BackgroundMobile stroke unit (MSU) is deployed to shorten the duration of ischemic stroke recognition to thrombolysis treatment, thus reducing disability, mortality after an acute stroke attack, and related economic burden. Therefore, we conducted a comprehensive systematic review of the clinical trial and economic literature focusing on various outcomes of MSU compared with conventional emergency medical services (EMS).MethodsAn electronic search was conducted in four databases (PubMed, OVID Medline, Embase, and the Cochrane Controlled Register of Trials) from 1990 to 2021. In these trials, patients with acute stroke were assigned to receive either MSU or EMS, with clinical and economic outcomes. First, we extracted interested data in the pooled population and conducted a subgroup analysis to examine related heterogeneity. We then implemented a descriptive analysis of economic outcomes. All analyses were performed with R 4.0.1 software.ResultsA total of 22,766 patients from 16 publications were included. In total 7,682 (n = 33.8%) were treated in the MSU and 15,084 (n = 66.2%) in the conventional EMS. Economic analysis were available in four studies, of which two were based on trial data and the others on model simulations. The pooled analysis of time metrics indicated a mean reduction of 32.64 min (95% confidence interval: 23.38–41.89, p < 0.01) and 28.26 minutes (95% CI: 16.11–40.41, p < 0.01) in the time-to-therapy and time-to-CT completion, respectively in the MSU. However, there was no significant difference on stroke-related neurological events (OR = 0.94, 95% CI: 0.70–1.27, p = 0.69) and in-hospital mortality (OR = 1.11, 95% CI: 0.83–1.50, p = 0.48) between the MSU and EMS. The proportion of patients with modified Ranking scale (mRS) of 0–2 at 90 days from onset was higher in the MSU than EMS (p < 0.05). MSU displayed favorable benefit-cost ratios (2.16–6.85) and incremental cost-effectiveness ratio ($31,911 /QALY and $38,731 per DALY) comparing to EMS in multiple economic publications. Total cost data based on 2014 USD showed that the MSU has the highest cost in Australia ($1,410,708) and the lowest cost in the USA ($783,463).ConclusionA comprehensive analysis of current research suggests that MUS, compared with conventional EMS, has a better performance in terms of time metrics, safety, long-term medical benefits, and cost-effectiveness.

The noise corruption problem commonly exists in hyperspectral images (HSIs) and severely affects the accuracy of hyperspectral unmixing algorithms. The noise formulation existing in HSIs is relatively complex and would change in conjunction with different devices and imaging settings. For real applications, applying denoising approaches without accurate close-to-reality noise modeling before unmixing may not improve, but rather degrade the unmixing performance. This study proposes a robust hyperspectral unmixing method with practical learning-based hyperspectral image denoising. We formulated a close-to-reality noise model for hyperspectral data and provide a calibration approach for the noise parameters. On the basis of the calibrated noise model, synthetic data were generated and used for training a KST-based denoising network. The noisy hyperspectral data were firstly denoised by the trained denoising network and were then used to perform the unmixing process. A variety of unmixing algorithms can be integrated into our method to improve the accuracy of unmixing in noisy situations. In the experiments, several widely used unmixing algorithms were employed to verify the effect of the proposed method. The experimental results on both synthetic and real demonstrated that our proposed method can handle HSI data with various gain settings and helps to improve the unmixing performance effectively.

The objective was to explore the function of gene differential expressions between lung cancer tissues and the interaction between the relevant encoded proteins, thereby analyzing the important genes closely related to lung cancer. A total of 120 samples from the GEO database (including two groups, i.e., 60 lung cancer in situ specimens and 60 normal specimens) were taken as the research objects, which were submitted to the analysis of signaling pathway, biological function enrichment, and protein interactions to reveal the molecular driving mechanism of lung cancer. Results: A total of 875 differentially expressed genes were obtained, including 291 up-regulated genes and 584 down-regulated genes. The up-regulated genes were mainly involved in biological processes such as protein metabolism, protein hydrolysis, mitosis, and cell division. Down-regulated genes were mainly involved in neutrophil chemotaxis, inflammatory response, immune response, and angiogenesis. The protein expression of high expression genes and low expression genes in patients were higher than those in the control group. The protein corresponding to the high expression gene was highly expressed in the patient group. Meanwhile, the proteins corresponding to the low expression genes were also expressed in the patient group, which showed that although the proteins corresponding to the low expression genes were low in the patients, they were still the target genes related to lung cancer. In conclusion, the molecular driving mechanism in lung cancer was mainly related to protein metabolism, proteolysis, mitosis, and cell division. It was found that TOP2A, CCNB1, CCNA2, CDK1, and TTK might be the critical target genes of lung cancer.

Background: During the COVID-19 pandemic, many patients admitted to hospital for treatment have recovered and been discharged; however, in some instances, these same patients are re-admitted due to a second fever or a positive COVID-19 PCR test result. To ascertain whether it is necessary to treat these patients in hospitals, especially in asymptomatic cases, we summarize and analyze the clinical and treatment characteristics of patients re-admitted to hospital with a second COVID-19 infection.Methods: Of the 141 COVID-19 cases admitted to the Wenzhou Central Hospital between January 17, 2020, to March 5, 2020, which were followed until March 30, 2020, 12 patients were re-admitted with a second COVID-19 infection. Data was collected and analyzed from their clinical records, lab indexes, commuted tomography (CT), and treatment strategies.Results: Most of the 141 patients had positive outcomes from treatment, with only 12 (8.5%) being re-admitted. In this sub-group: one (8.3%) had a fever, a high white blood cell count (WBC), and progressive CT changes; and one (8.3%) had increased transaminase. The PCR tests of these two patients returned negative results. Another 10 patients were admitted due to a positive PCR test result, seven of which were clinically asymptomatic. Compared to the CT imaging following their initial discharge, the CT imaging of all patients was significantly improved, and none required additional oxygen or mechanical ventilation during their second course of treatment.Conclusions: The prognoses of the re-admitted patients were good with no serious cases. We conclude that home treatment with concentrated medical observation is a safe and feasible course of treatment if the patient returns a positive PCR test result but does not display serious clinical symptoms. During medical observation, patients with underlying conditions should remain a primary focus, but most do not need to be re-admitted to the hospital.

This paper presents a new parameterized nonlinear least squares (PNLS) algorithm for unsupervised nonlinear spectral unmixing (UNSU). The PNLS-based algorithms transform the original optimization problem with respect to the endmembers, abundances, and nonlinearity coefficients estimation into separate alternate parameterized nonlinear least squares problems. Owing to the Sigmoid parameterization, the PNLS-based algorithms are able to thoroughly relax the additional nonnegative constraint and the nonnegative constraint in the original optimization problems, which facilitates finding a solution to the optimization problems . Subsequently, we propose to solve the PNLS problems based on the Gauss–Newton method. Compared to the existing nonnegative matrix factorization (NMF)-based algorithms for UNSU, the well-designed PNLS-based algorithms have faster convergence speed and better unmixing accuracy. To verify the performance of the proposed algorithms, the PNLS-based algorithms and other state-of-the-art algorithms are applied to synthetic data generated by the Fan model and the generalized bilinear model (GBM), as well as real hyperspectral data. The results demonstrate the superiority of the PNLS-based algorithms.

Hyperspectral unmixing aims to estimate a set of endmembers and corresponding abundances in pixels. Nonnegative matrix factorization (NMF) and its extensions with various constraints have been widely applied to hyperspectral unmixing. L 1 / 2 and L 2 regularizers can be added to NMF to enforce sparseness and evenness, respectively. In practice, a region in a hyperspectral image may possess different sparsity levels across locations. The problem remains as to how to impose constraints accordingly when the level of sparsity varies. We propose a novel nonnegative matrix factorization with data-guided constraints (DGC-NMF). The DGC-NMF imposes on the unknown abundance vector of each pixel with either an L 1 / 2 constraint or an L 2 constraint according to its estimated mixture level. Experiments on the synthetic data and real hyperspectral data validate the proposed algorithm.