MMP12 is a Potential Predictive and Prognostic Biomarker of Various Cancers Including Lung Adenocarcinoma.

OBJECTIVE: This study sought to explore the clinical value of matrix metalloproteinases 12 (MMP12) in multiple cancers, including lung adenocarcinoma (LUAD). METHODS: Using >10,000 samples, this retrospective study demonstrated the first pan-cancer analysis of MMP12. The expression of MMP12 between cancer groups and their control groups was analyzed using Wilcoxon rank-sum tests. The clinical significance of MMP12 expression in multiple cancers was assessed using receiver operating characteristic curves, Kaplan-Meier curves, and univariate Cox analysis. A further LUAD-related analysis based on 4565 multi-center and in-house samples was performed to verify the findings regarding MMP12 in pan-cancer analysis partly. RESULTS: MMP12 mRNA is highly expressed in 13 cancers compared to their controls, and the MMP12 protein level is elevated in some of these cancers (e.g., colon adenocarcinoma) (P < .05). MMP12 expression makes it feasible to distinguish 21 cancer tissues from normal tissues (AUC = 0.86). A high MMP12 expression is a prognosis risk factor in eight cancers, such as adrenocortical carcinoma (hazard ratio >1, P < .05). The elevated MMP12 expression is also a prognosis protective factor in breast-invasive carcinoma and colon adenocarcinoma (hazard ratio <1, P < .05). Some pan-cancer findings regarding MMP12 are verified in LUAD-MMP12 expression is upregulated in LUAD at both the mRNA and protein levels (P < .05), has the potential to distinguish LUAD with considerable accuracy (AUC = .91), and plays a risk prognosis factor for patients with the disease (P < .05). CONCLUSIONS: MMP12 is highly expressed in most cancers and may serve as a novel biomarker for the prediction and prognosis of numerous cancers.

CDK6 is a novel predictive and prognosis biomarker correlated with immune infiltrates in multiple human neoplasms, including small cell lung carcinoma.

The roles of cyclin-dependent kinase 6 (CDK6) in various cancers, including small cell lung carcinoma (SCLC), remain unclear. Here, 111,54 multi-center samples were investigated to determine the expression, clinical significance, and underlying mechanisms of CDK6 in 34 cancers. The area under the curve (AUC), Cox regression analysis, and the Kaplan-Meier curves were used to explore the clinical value of CDK6 in cancers. Gene set enrichment analysis and correlation analysis were performed to detect potential CDK6 mechanisms. CDK6 expression was essential in 24 cancer cell types. Abnormal CDK6 expression was observed in 14 cancer types (e.g., downregulated in breast invasive carcinoma; p < 0.05). CDK6 allowed six cancers to be distinguished from their controls (AUC > 0.750). CDK6 expression was a prognosis marker for 13 cancers (e.g., adrenocortical carcinoma; p < 0.05). CDK6 was correlated with several immune-related signaling pathways and the infiltration levels of certain immune cells (e.g., CD8+ T cells; p < 0.05). Downregulated CDK6 mRNA and protein levels were observed in SCLC (p < 0.05, SMD = - 0.90). CDK6 allowed the identification of SCLC status (AUC = 0.91) and predicted a favorable prognosis for SCLC patients (p < 0.05). CDK6 may be a novel biomarker for the prediction and prognosis of several cancers, including SCLC.

ITGB4 Serves as an Identification and Prognosis Marker Associated with Immune Infiltration in Small Cell Lung Carcinoma.

Integrin beta 4 (ITGB4) is a vital factor for numerous cancers. However, no reports regarding ITGB4 in small cell lung carcinoma (SCLC) have been found in the existing literature. This study systematically investigated the expression and clinical value of ITGB4 in SCLC using multi-center and large-sample (n = 963) data. The ITGB4 expression levels between SCLC and control tissues were compared using standardized mean difference and Wilcoxon rank-sum test. The clinical significance of the gene in SCLC was observed using Cox regression and Kaplan-Meier curves. ITGB4 is overexpressed in multiple cancers and represents significant value in distinguishing among cancer samples (AUC = 0.91) and predicting the prognoses (p < 0.05) of patients with different cancers. In contrast, decreased ITGB4 mRNA expression was determined in SCLC (SMD < 0), and this finding was further confirmed at protein levels using in-house specimens (p < 0.05). This decrease in expression may be attributed to the regulatory role of estrogen receptor 1. ITGB4 may participate in the progression of SCLC by affecting several signaling pathways (e.g., tumor necrosis factor signaling pathway) and a series of immune cells (e.g., dendritic cells) (p < 0.05). The gene may serve as a potential marker for predicting the disease status (AUC = 0.97) and prognoses (p < 0.05) of patients with SCLC. Collectively, ITGB4 was identified as an identification and prognosis marker associated with immune infiltration in SCLC.

Influence of Si and SiC Coating on the Microstructures and Mechanical Properties of C/C Bolts.

High-performance bolts made of carbon/carbon (C/C) composites are necessary for connecting thermally-insulating structural components of aerospace vehicles. To enhance the mechanical properties of the C/C bolt, a new silicon infiltration-modified C/C (C/C-SiC) bolt was developed via vapor silicon infiltration. The effects of silicon infiltration on microstructure and mechanical properties were systematically studied. Findings reveal that dense and uniform SiC-Si coating has been formed after silicon infiltration of the C/C bolt, strongly bonding with the C matrix. Under tensile stress, the C/C-SiC bolt undergoes a tensile failure of studs, while the C/C bolt is subject to the pull-out failure of threads. The breaking strength of the former (55.16 MPa) is 26.83% higher than the failure strength of the latter (43.49 MPa). Under double-sided shear stress, both the crushing of threads and the shear failure of studs occur within two bolts. As a result, the shear strength of the former (54.73 MPa) exceeds that of the latter (43.88 MPa) by 24.73%. According to CT and SEM analysis, matrix fracture, fiber debonding, and fiber bridging are the main failure modes. Therefore, a mixed coating formed by silicon infiltration can effectively transfer loads from coating to carbon matrix and carbon fiber, thereby enhancing the load-bearing capacity of C/C bolts.

Automatic mammographic breast density classification in Chinese women: clinical validation of a deep learning model.

BACKGROUND: High breast density is a strong risk factor for breast cancer. As such, high consistency and accuracy in breast density assessment is necessary. PURPOSE: To validate our proposed deep learning (DL) model and explore its impact on radiologists on density assessments. MATERIAL AND METHODS: A total of 3732 mammographic cases were collected as a validated set: 1686 cases before the implementation of the DL model and 2046 cases after the DL model. Five radiologists were divided into two groups (junior and senior groups) to assess all mammograms using either two- or four-category evaluation. Linear-weighted kappa (K) and intraclass correlation coefficient (ICC) statistics were used to analyze the consistency between radiologists before and after implementation of the DL model. RESULTS: The accuracy and clinical acceptance of the DL model for the junior group were 96.3% and 96.8% for two-category evaluation, and 85.6% and 89.6% for four-category evaluation, respectively. For the senior group, the accuracy and clinical acceptance were 95.5% and 98.0% for two-category evaluation, and 84.3% and 95.3% for four-category evaluation, respectively. The consistency within the junior group, the senior group, and among all radiologists improved with the help of the DL model. For two-category, their K and ICC values improved to 0.81, 0.81, and 0.80 from 0.73, 0.75, and 0.76. And for four-category, their K and ICC values improved to 0.81, 0.82, and 0.82 from 0.73, 0.79, and 0.78, respectively. CONCLUSION: The DL model showed high accuracy and clinical acceptance in breast density categories. It is helpful to improve radiologists' consistency.

Downregulation of zinc finger protein 71 in laryngeal squamous cell carcinoma tissues and its potential molecular mechanism and clinical significance: a study based on immunohistochemistry staining and data mining.

BACKGROUND: The molecular mechanism of laryngeal squamous cell carcinoma (LSCC) is not completely clear, which leads to poor prognosis and treatment difficulties for LSCC patients. To date, no study has reported the exact expression level of zinc finger protein 71 (ZNF71) and its molecular mechanism in LSCC. METHODS: In-house immunohistochemistry (IHC) staining (33 LSCC samples and 29 non-LSCC samples) was utilized in analyzing the protein expression level of ZNF71 in LSCC. Gene chips and high-throughput sequencing data collected from multiple public resources (313 LSCC samples and 192 non-LSCC samples) were utilized in analyzing the exact mRNA expression level of ZNF71 in LSCC. Single-cell RNA sequencing (scRNA-seq) data was used to explore the expression status of ZNF71 in different LSCC subpopulations. Enrichment analysis of ZNF71, its positively and differentially co-expressed genes (PDCEGs), and its downstream target genes was employed to detect the potential molecular mechanism of ZNF71 in LSCC. Moreover, we conducted correlation analysis between ZNF71 expression and immune infiltration. RESULTS: ZNF71 was downregulated at the protein level (area under the curve [AUC] = 0.93, p < 0.0001) and the mRNA level (AUC = 0.71, p = 0.023) in LSCC tissues. Patients with nodal metastasis had lower protein expression level of ZNF71 than patients without nodal metastasis (p < 0.05), and male LSCC patients had lower mRNA expression level of ZNF71 than female LSCC patients (p < 0.01). ZNF71 was absent in different LSCC subpopulations, including cancer cells, plasma cells, and tumor-infiltrated immune cells, based on scRNA-seq analysis. Enrichment analysis showed that ZNF71 and its PDCEGs may influence the progression of LSCC by regulating downstream target genes of ZNF71. These downstream target genes of ZNF71 were mainly enriched in tight junctions. Moreover, downregulation of ZNF71 may influence the development and even therapy of LSCC by reducing immune infiltration. CONCLUSION: Downregulation of ZNF71 may promote the progression of LSCC by reducing tight junctions and immune infiltration; this requires further study.

Downregulation of zinc finger protein 71 expression in oral squamous cell carcinoma tissues and its underlying molecular mechanism.

BACKGROUND: Patients with oral squamous cell carcinoma (OSCC) have poor prognoses due to a limited understanding of the pathogenesis of OSCC. Zinc finger protein (ZNF) is the largest transcription factor family in the human genome and exert diverse and important functions. Nevertheless, the exact expression status and molecular mechanism of ZNF71 have not been described in OSCC. Therefore, this study aimed to identify the specific expression level of ZNF71 in OSCC tissues and to further interpret the potential molecular mechanism of ZNF71 in the pathogenesis of OSCC. METHODS: In-house immunohistochemical staining of 116 OSCC samples and 29 non-OSCC samples was employed to detect the expression status of ZNF71 at the protein level of OSCC tissues. Single-cell RNA sequencing data from 7 OSCC samples was used to explore the expression landscape of ZNF71 in different cell types from OSCC tissues. High-throughput RNA sequencing data and gene chips data from 893 OSCC samples and 301 non-OSCC samples were utilized to identify the specific expression level of ZNF71 at the bulk mRNA level of OSCC tissues. Here, standardized mean difference (SMD) value was applied to calculate the expression differences between OSCC group and non-OSCC group. Multiple datasets were included; hence, the results were considered to be more reliable. Sensitivity analysis was conducted to evaluate the stability of the results. Enrichment analysis and immune infiltration analysis were used to explore the underlying molecular mechanism of ZNF71 in OSCC. RESULTS: ZNF71 was significantly downregulated in OSCC tissues at the protein level (SMD = -1.96, 95 % confidence interval [95 % CI]: -2.43 to -1.50). ZNF71 was absent in various cell types from OSCC tissues including cancerous epithelial cells and tumor-infiltrating immune cells. ZNF71 was downregulated in OSCC tissues at the bulk mRNA level (SMD = -0.38, 95 % CI: -0.75 to -0.02). Enrichment analysis showed that positively and differentially co-expressed genes mainly concentrated on "herpes simplex virus 1 infection" and "regulation of plasma membrane bounded cell projection organization", and negatively and differentially co-expressed genes mainly participated in "cell cycle" and "DNA metabolic process". Moreover, the putative target genes of ZNF71 mainly participated in "cellular respiration" and "protein catabolic process". Finally, immune infiltration analysis revealed that ZNF71 expression was positively correlated with multiple immune cells including activated B cells, memory B cells, and natural killer (NK) cells, and negatively correlated with various immune cells, including CD56 bright NK cells, neutrophil, and immature dendritic cells. CONCLUSION: The downregulation of ZNF71 may influence the initiation and promotion of OSCC by reducing immune infiltration, accelerating cell cycle progression, and affecting metabolic process, and this requires further research.

Downregulation of MicroRNA-1 and Its Potential Molecular Mechanism in Nasopharyngeal Cancer: An Investigation Combined with In Silico and In-House Immunohistochemistry Validation.

BACKGROUND: This study was aimed at elucidating the molecular biological mechanisms of microRNA-1 (miR-1) in nasopharyngeal carcinoma (NPC). METHOD: In this study, we performed a pooled analysis of miR-1 expression data derived from public databases, such as GEO, ArrayExpress, TCGA, and GTEx. The miRWalk 2.0 database, combined with the mRNA microarray datasets, was used to screen the target genes, and the genes were then subjected to Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analysis using the DAVID 6.8 database. We then used the STRING 11.0 database and Cytoscape 3.80 software to construct a protein-protein interaction (PPI) network for screening hub genes. Immunohistochemistry (IHC) was further used to validate the expression of hub genes. Finally, potential therapeutic agents for NPC were screened by the Connectivity Map (cMap) database. RESULTS: Pooled analysis showed that miR-1 expression was significantly decreased in NPC (SMD = -0.57; P < 0.05). The summary receiver operating characteristic curve suggested that miR-1 had a good ability to distinguish cancerous tissues from noncancerous tissues (AUC = 0.78). The results of GO analysis focused on mitotic nuclear division, DNA replication, cell division, cell adhesion, extracellular space, kinesin complex, and extracellular matrix (ECM) structural constituent. The KEGG analysis suggested that the target genes played a role in key signaling pathways, such as cell cycle, focal adhesion, cytokine-cytokine receptor interaction, ECM-receptor interaction, and PI3K/Akt signaling pathway. The PPI network suggested that cyclin-dependent kinase 1 (CDK1) was the hub gene, and the CDK1 protein was subsequently confirmed to be significantly upregulated in NPC tissues by IHC. Finally, potential therapeutic drugs, such as masitinib, were obtained by the cMap database. CONCLUSION: miR-1 may play a vital part in NPC tumorigenesis and progression by regulating focal adhesion kinase to participate in cell mitosis, regulating ECM degradation, and affecting the PI3K/Akt signaling pathway. miR-1 has the potential to be a therapeutic target for NPC.

The added value of an artificial intelligence system in assisting radiologists on indeterminate BI-RADS 0 mammograms.

OBJECTIVES: To investigate the value of an artificial intelligence (AI) system in assisting radiologists to improve the assessment accuracy of BI-RADS 0 cases in mammograms. METHODS: We included 34,654 consecutive digital mammography studies, collected between January 2011 and January 2019, among which, 1088 cases from 1010 unique patients with initial BI-RADS 0 assessment who were recalled during 2 years of follow-up were used in this study. Two mid-level radiologists retrospectively re-assessed these BI-RADS 0 cases with the assistance of an AI system developed by us previously. In addition, four entry-level radiologists were split into two groups to cross-read 80 cases with and without the AI. Diagnostic performance was evaluated using the follow-up diagnosis or biopsy results as the reference standard. RESULTS: Of the 1088 cases, 626 were actually normal (BI-RADS 1 and no recall required). Assisted by the AI system, 351 (56%) and 362 (58%) normal cases were correctly identified by the two mid-level radiologists hence can be avoided for unnecessary follow-ups. However, they would have missed 12 (10 invasive cancers and 2 ductal carcinoma in situ cancers) and 6 (invasive cancers) malignant lesions respectively as a result. These missed lesions were not highly malignant tumors. The inter-rater reliability of entry-level radiologists increased from 0.20 to 0.30 (p < 0.005) by introducing the AI. CONCLUSION: The AI system can effectively assist mid-level radiologists in reducing unnecessary follow-ups of mammographically indeterminate breast lesions and reducing the benign biopsy rate without missing highly malignant tumors. KEY POINTS: • The artificial intelligence system could assist mid-level radiologists in effectively reducing unnecessary BI-RADS 0 mammogram recalls and the benign biopsy rate without missing highly malignant tumors. • The artificial intelligence system was capable of detecting low suspicion lesions from heterogeneously and extremely dense breasts that radiologists tended to miss. • The use of an artificial intelligence system may improve the inter-rater reliability and sensitivity, and reduce the reading time of entry-level radiologists in assessing potential lesions in BI-RADS 0 mammograms.

MommiNet-v2: Mammographic multi-view mass identification networks.

Many existing approaches for mammogram analysis are based on single view. Some recent DNN-based multi-view approaches can perform either bilateral or ipsilateral analysis, while in practice, radiologists use both to achieve the best clinical outcome. MommiNet is the first DNN-based tri-view mass identification approach, which can simultaneously perform bilateral and ipsilateral analysis of mammographic images, and in turn, can fully emulate the radiologists' reading practice. In this paper, we present MommiNet-v2, with improved network architecture and performance. Novel high-resolution network (HRNet)-based architectures are proposed to learn the symmetry and geometry constraints, to fully aggregate the information from all views for accurate mass detection. A multi-task learning scheme is adopted to incorporate both Breast Imaging-Reporting and Data System (BI-RADS) and biopsy information to train a mass malignancy classification network. Extensive experiments have been conducted on the public DDSM (Digital Database for Screening Mammography) dataset and our in-house dataset, and state-of-the-art results have been achieved in terms of mass detection accuracy. Satisfactory mass malignancy classification result has also been obtained on our in-house dataset.

COVID-19-CT-CXR: A Freely Accessible and Weakly Labeled Chest X-Ray and CT Image Collection on COVID-19 From Biomedical Literature.

The latest threat to global health is the COVID-19 outbreak. Although there exist large datasets of chest X-rays (CXR) and computed tomography (CT) scans, few COVID-19 image collections are currently available due to patient privacy. At the same time, there is a rapid growth of COVID-19-relevant articles in the biomedical literature, including those that report findings on radiographs. Here, we present COVID-19-CT-CXR, a public database of COVID-19 CXR and CT images, which are automatically extracted from COVID-19-relevant articles from the PubMed Central Open Access (PMC-OA) Subset. We extracted figures, associated captions, and relevant figure descriptions in the article and separated compound figures into subfigures. Because a large portion of figures in COVID-19 articles are not CXR or CT, we designed a deep-learning model to distinguish them from other figure types and to classify them accordingly. The final database includes 1,327 CT and 263 CXR images (as of May 9, 2020) with their relevant text. To demonstrate the utility of COVID-19-CT-CXR, we conducted four case studies. (1) We show that COVID-19-CT-CXR, when used as additional training data, is able to contribute to improved deep-learning (DL) performance for the classification of COVID-19 and non-COVID-19 CT. (2) We collected CT images of influenza, another common infectious respiratory illness that may present similarly to COVID-19, and fine-tuned a baseline deep neural network to distinguish a diagnosis of COVID-19, influenza, or normal or other types of diseases on CT. (3) We fine-tuned an unsupervised one-class classifier from non-COVID-19 CXR and performed anomaly detection to detect COVID-19 CXR. (4) From text-mined captions and figure descriptions, we compared 15 clinical symptoms and 20 clinical findings of COVID-19 versus those of influenza to demonstrate the disease differences in the scientific publications. Our database is unique, as the figures are retrieved along with relevant text with fine-grained descriptions, and it can be extended easily in the future. We believe that our work is complementary to existing resources and hope that it will contribute to medical image analysis of the COVID-19 pandemic. The dataset, code, and DL models are publicly available at https://github.com/ncbi-nlp/COVID-19-CT-CXR.

Discriminative ensemble learning for few-shot chest x-ray diagnosis.

Few-shot learning is an almost unexplored area in the field of medical image analysis. We propose a method for few-shot diagnosis of diseases and conditions from chest x-rays using discriminative ensemble learning. Our design involves a CNN-based coarse-learner in the first step to learn the general characteristics of chest x-rays. In the second step, we introduce a saliency-based classifier to extract disease-specific salient features from the output of the coarse-learner and classify based on the salient features. We propose a novel discriminative autoencoder ensemble to design the saliency-based classifier. The classification of the diseases is performed based on the salient features. Our algorithm proceeds through meta-training and meta-testing. During the training phase of meta-training, we train the coarse-learner. However, during the training phase of meta-testing, we train only the saliency-based classifier. Thus, our method is first-of-its-kind where the training phase of meta-training and the training phase of meta-testing are architecturally disjoint, making the method modular and easily adaptable to new tasks requiring the training of only the saliency-based classifier. Experiments show as high as ∼19% improvement in terms of F1 score compared to the baseline in the diagnosis of chest x-rays from publicly available datasets.

Learning From Multiple Datasets With Heterogeneous and Partial Labels for Universal Lesion Detection in CT.

Large-scale datasets with high-quality labels are desired for training accurate deep learning models. However, due to the annotation cost, datasets in medical imaging are often either partially-labeled or small. For example, DeepLesion is such a large-scale CT image dataset with lesions of various types, but it also has many unlabeled lesions (missing annotations). When training a lesion detector on a partially-labeled dataset, the missing annotations will generate incorrect negative signals and degrade the performance. Besides DeepLesion, there are several small single-type datasets, such as LUNA for lung nodules and LiTS for liver tumors. These datasets have heterogeneous label scopes, i.e., different lesion types are labeled in different datasets with other types ignored. In this work, we aim to develop a universal lesion detection algorithm to detect a variety of lesions. The problem of heterogeneous and partial labels is tackled. First, we build a simple yet effective lesion detection framework named Lesion ENSemble (LENS). LENS can efficiently learn from multiple heterogeneous lesion datasets in a multi-task fashion and leverage their synergy by proposal fusion. Next, we propose strategies to mine missing annotations from partially-labeled datasets by exploiting clinical prior knowledge and cross-dataset knowledge transfer. Finally, we train our framework on four public lesion datasets and evaluate it on 800 manually-labeled sub-volumes in DeepLesion. Our method brings a relative improvement of 49% compared to the current state-of-the-art approach in the metric of average sensitivity. We have publicly released our manual 3D annotations of DeepLesion online.1 1https://github.com/viggin/DeepLesion_manual_test_set.

A disentangled generative model for disease decomposition in chest X-rays via normal image synthesis.

The interpretation of medical images is a complex cognition procedure requiring cautious observation, precise understanding/parsing of the normal body anatomies, and combining knowledge of physiology and pathology. Interpreting chest X-ray (CXR) images is challenging since the 2D CXR images show the superimposition on internal organs/tissues with low resolution and poor boundaries. Unlike previous CXR computer-aided diagnosis works that focused on disease diagnosis/classification, we firstly propose a deep disentangled generative model (DGM) simultaneously generating abnormal disease residue maps and "radiorealistic" normal CXR images from an input abnormal CXR image. The intuition of our method is based on the assumption that disease regions usually superimpose upon or replace the pixels of normal tissues in an abnormal CXR. Thus, disease regions can be disentangled or decomposed from the abnormal CXR by comparing it with a generated patient-specific normal CXR. DGM consists of three encoder-decoder architecture branches: one for radiorealistic normal CXR image synthesis using adversarial learning, one for disease separation by generating a residue map to delineate the underlying abnormal region, and the other one for facilitating the training process and enhancing the model's robustness on noisy data. A self-reconstruction loss is adopted in the first two branches to enforce the generated normal CXR image to preserve similar visual structures as the original CXR. We evaluated our model on a large-scale chest X-ray dataset. The results show that our model can generate disease residue/saliency maps (coherent with radiologist annotations) along with radiorealistic and patient specific normal CXR images. The disease residue/saliency map can be used by radiologists to improve the CXR reading efficiency in clinical practice. The synthesized normal CXR can be used for data augmentation and normal control of personalized longitudinal disease study. Furthermore, DGM quantitatively boosts the diagnosis performance on several important clinical applications, including normal/abnormal CXR classification, and lung opacity classification/detection.

Subcellular localization of the porcine deltacoronavirus nucleocapsid protein.

Porcine deltacoronavirus (PDCoV) has been recently identified as an emerging enteropathogenic coronavirus that mainly infects newborn piglets and causes enteritis, diarrhea and high mortality. Although coronavirus N proteins have multifarious activities, the subcellular localization of the PDCoV N protein is still unknown. Here, we produced mouse monoclonal antibodies against the PDCoV N protein. Experiments using anti-haemagglutinin antibodies and these monoclonal antibodies revealed that the PDCoV N protein is shuttled into the nucleolus in both ectopic PDCoV N-expressing cells and PDCoV-infected cells. The results of deletion mutagenesis experiments demonstrated that the predicted nucleolar localization signal at amino acids 295-318 is critical for nucleolar localization. Cumulatively, our study yielded a monoclonal antibody against the PDCoV N protein and revealed a mechanism by which the PDCoV N protein translocated into the nucleolus. The tolls and findings from this work will facilitate further investigations on the functions of the PDCoV N protein.

COVID-19-CT-CXR: a freely accessible and weakly labeled chest X-ray and CT image collection on COVID-19 from biomedical literature.

The latest threat to global health is the COVID-19 outbreak. Although there exist large datasets of chest X-rays (CXR) and computed tomography (CT) scans, few COVID-19 image collections are currently available due to patient privacy. At the same time, there is a rapid growth of COVID-19-relevant articles in the biomedical literature. Here, we present COVID-19-CT-CXR, a public database of COVID-19 CXR and CT images, which are automatically extracted from COVID-19-relevant articles from the PubMed Central Open Access (PMC-OA) Subset. We extracted figures, associated captions, and relevant figure descriptions in the article and separated compound figures into subfigures. We also designed a deep-learning model to distinguish them from other figure types and to classify them accordingly. The final database includes 1,327 CT and 263 CXR images (as of May 9, 2020) with their relevant text. To demonstrate the utility of COVID-19-CT-CXR, we conducted four case studies. (1) We show that COVID-19-CT-CXR, when used as additional training data, is able to contribute to improved DL performance for the classification of COVID-19 and non-COVID-19 CT. (2) We collected CT images of influenza and trained a DL baseline to distinguish a diagnosis of COVID-19, influenza, or normal or other types of diseases on CT. (3) We trained an unsupervised one-class classifier from non-COVID-19 CXR and performed anomaly detection to detect COVID-19 CXR. (4) From text-mined captions and figure descriptions, we compared clinical symptoms and clinical findings of COVID-19 vs. those of influenza to demonstrate the disease differences in the scientific publications. We believe that our work is complementary to existing resources and hope that it will contribute to medical image analysis of the COVID-19 pandemic. The dataset, code, and DL models are publicly available at https://github.com/ncbi-nlp/COVID-19-CT-CXR.

Automated abnormality classification of chest radiographs using deep convolutional neural networks.

As one of the most ubiquitous diagnostic imaging tests in medical practice, chest radiography requires timely reporting of potential findings and diagnosis of diseases in the images. Automated, fast, and reliable detection of diseases based on chest radiography is a critical step in radiology workflow. In this work, we developed and evaluated various deep convolutional neural networks (CNN) for differentiating between normal and abnormal frontal chest radiographs, in order to help alert radiologists and clinicians of potential abnormal findings as a means of work list triaging and reporting prioritization. A CNN-based model achieved an AUC of 0.9824 ± 0.0043 (with an accuracy of 94.64 ± 0.45%, a sensitivity of 96.50 ± 0.36% and a specificity of 92.86 ± 0.48%) for normal versus abnormal chest radiograph classification. The CNN model obtained an AUC of 0.9804 ± 0.0032 (with an accuracy of 94.71 ± 0.32%, a sensitivity of 92.20 ± 0.34% and a specificity of 96.34 ± 0.31%) for normal versus lung opacity classification. Classification performance on the external dataset showed that the CNN model is likely to be highly generalizable, with an AUC of 0.9444 ± 0.0029. The CNN model pre-trained on cohorts of adult patients and fine-tuned on pediatric patients achieved an AUC of 0.9851 ± 0.0046 for normal versus pneumonia classification. Pretraining with natural images demonstrates benefit for a moderate-sized training image set of about 8500 images. The remarkable performance in diagnostic accuracy observed in this study shows that deep CNNs can accurately and effectively differentiate normal and abnormal chest radiographs, thereby providing potential benefits to radiology workflow and patient care.

Visual and Semantic Knowledge Transfer for Large Scale Semi-Supervised Object Detection.

Deep CNN-based object detection systems have achieved remarkable success on several large-scale object detection benchmarks. However, training such detectors requires a large number of labeled bounding boxes, which are more difficult to obtain than image-level annotations. Previous work addresses this issue by transforming image-level classifiers into object detectors. This is done by modeling the differences between the two on categories with both image-level and bounding box annotations, and transferring this information to convert classifiers to detectors for categories without bounding box annotations. We improve this previous work by incorporating knowledge about object similarities from visual and semantic domains during the transfer process. The intuition behind our proposed method is that visually and semantically similar categories should exhibit more common transferable properties than dissimilar categories, e.g. a better detector would result by transforming the differences between a dog classifier and a dog detector onto the cat class, than would by transforming from the violin class. Experimental results on the challenging ILSVRC2013 detection dataset demonstrate that each of our proposed object similarity based knowledge transfer methods outperforms the baseline methods. We found strong evidence that visual similarity and semantic relatedness are complementary for the task, and when combined notably improve detection, achieving state-of-the-art detection performance in a semi-supervised setting.

Long non-coding RNA reprogramming (lncRNA-ROR) regulates cell apoptosis and autophagy in chondrocytes.

Long Non-Coding RNA Reprogramming (lncRNA-ROR) plays an important role in regulating various biologic processes, whereas the effect of lncRNA-ROR in osteoarthritis (OA) is little studied. This study aimed to explore lncRNA-ROR expression in articular cartilage and identify the functional mechanism of lncRNA-ROR in OA. OA cartilage tissues were obtained from 15 OA patients, and 6 normal cartilage tissues were set as controls. Chondrocytes were isolated from the collected cartilage tissues. lncRNA-ROR was knockdown in normal cells and overexpressed in OA cells. Cell viability was determined with Cell Counting Kit-8 assay, and apoptosis was measured using flow cytometric analysis. Moreover, proteins and mRNAs involved in this study were also measured using Western blotting and quantitative real-time PCR (qPCR). Level of lncRNA-ROR was decreased in OA compared with normal chondrocytes, and overexpression of lncRNA-ROR dramatically promoted cell viability of OA chondrocytes. In addition, knockdown lncRNA-ROR inhibited apoptosis and promoted autophagy of normal chondrocytes. Moreover, lncRNA-ROR inhibited the expression of p53 in both mRNA and protein levels. Furthermore, we revealed that lncRNA-ROR regulated apoptosis and autophagy of chondrocytes via HIF1α and p53. The results indicated that lncRNA-ROR played a critical role in the pathogenesis of OA, suggesting that lncRNA-ROR could serve as a new potential therapeutic target for OA.

Down-regulation of microRNA-23b aggravates LPS-induced inflammatory injury in chondrogenic ATDC5 cells by targeting PDCD4.

OBJECTIVES: Osteoarthritis (OA), characterized by degradation of articular cartilage, is a leading cause of disability. As the only cell type present in cartilage, chondrocytes play curial roles in the progression of OA. In our study, we aimed to explore the roles of miR-23b in the lipopolysaccharide (LPS)-induced inflammatory injury. MATERIALS AND METHODS: LPS-induced cell injury of ATDC5 cells was evaluated by the loss of cell viability, enhancement of cell apoptosis, alteration of apoptosis-associated proteins, and release of inflammatory cytokines. Then, miR-23b level after LPS treatment was assessed by qRT-PCR. Next, the effects of aberrantly expressed miR-23b on the LPS-induced inflammatory injury were explored. The possible target genes of miR-23b were virtually screened by informatics and verified by luciferase assay. Subsequently, whether miR-23b functioned through regulating the target gene was validated. The involved signaling pathways were investigated finally. RESULTS: Cell viability was decreased but cell apoptosis, as well as release of inflammatory cytokines, was enhanced by LPS treatment. MiR-23b was down-regulated by LPS and its overexpression alleviated LPS-induced inflammatory injury. PDCD4, negatively regulated by miR-23b expression, was verified as a target gene of miR-23b. Following experiments showed miR-23b alleviated LPS-induced cell injury through down-regulating PDCD4 expression. Phosphorylated levels of key kinases in the NF-κB pathway, as well as expressions of key kinases in the Notch pathways, were increased by PDCD4 overexpression. CONCLUSION: MiR-23b was down-regulated after LPS treatment, and its overexpression ameliorated LPS-induced inflammatory injury in ATDC5 cells by targeting PDCD4, which could activate the NF-κB/Notch pathways.