Application of Deep Learning in Generating Structured Radiology Reports: A Transformer-Based Technique.

Since radiology reports needed for clinical practice and research are written and stored in free-text narrations, extraction of relative information for further analysis is difficult. In these circumstances, natural language processing (NLP) techniques can facilitate automatic information extraction and transformation of free-text formats to structured data. In recent years, deep learning (DL)-based models have been adapted for NLP experiments with promising results. Despite the significant potential of DL models based on artificial neural networks (ANN) and convolutional neural networks (CNN), the models face some limitations to implement in clinical practice. Transformers, another new DL architecture, have been increasingly applied to improve the process. Therefore, in this study, we propose a transformer-based fine-grained named entity recognition (NER) architecture for clinical information extraction. We collected 88 abdominopelvic sonography reports in free-text formats and annotated them based on our developed information schema. The text-to-text transfer transformer model (T5) and Scifive, a pre-trained domain-specific adaptation of the T5 model, were applied for fine-tuning to extract entities and relations and transform the input into a structured format. Our transformer-based model in this study outperformed previously applied approaches such as ANN and CNN models based on ROUGE-1, ROUGE-2, ROUGE-L, and BLEU scores of 0.816, 0.668, 0.528, and 0.743, respectively, while providing an interpretable structured report.

EfficientMask-Net for face authentication in the era of COVID-19 pandemic.

Today, we are facing the COVID-19 pandemic. Accordingly, properly wearing face masks has become vital as an effective way to prevent the rapid spread of COVID-19. This research develops an Efficient Mask-Net method for low-power devices, such as mobile and embedding models with low-memory requirements. The method identifies face mask-wearing conditions in two different schemes: I. Correctly Face Mask (CFM), Incorrectly Face Mask (IFM), and Not Face Mask (NFM) wearing; II. Uncovered Chin IFM, Uncovered Nose IFM, and Uncovered Nose and Mouth IFM. The proposed method can also be helpful to unmask the face for face authentication based on unconstrained 2D facial images in the wild. In this study, deep convolutional neural networks (CNNs) were employed as feature extractors. Then, deep features were fed to a recently proposed large margin piecewise linear (LMPL) classifier. In the experimental study, lightweight and very powerful mobile implementation of CNN models were evaluated, where the novel "EffientNetb0" deep feature extractor with LMPL classifier outperformed well-known end-to-end CNN models, as well as conventional image classification methods. It achieved high accuracies of 99.53 and 99.64% in fulfilling the two mentioned tasks, respectively.

Organ-specific or personalized treatment for COVID-19: rationale, evidence, and potential candidates.

Although extrapulmonary manifestations of coronavirus disease 2019 (COVID-19) are increasingly reported, no effective therapeutic strategy for these multisystemic complications is available due to a poor understanding of the pathophysiology of COVID-19 multiorgan involvement. In this study, differentially expressed genes (DEGs) of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-infected extrapulmonary organs including human pluripotent stem cells (hPSCs)-derived liver organoids and choroid plexus organoids besides transformed lung alveolar (A549) cells were analyzed. First, pathway enrichment analysis was done to compare the underlying biological pathways enriched upon SARS-CoV-2 infection in different organs. Then, these lists of DEGs were used in a connectivity map (CMap)-based drug repurposing experiment. Also, protein-protein interaction (PPI) network analysis was done to compare the associated hub genes. The results revealed different biological pathways and genes responsible for SARS-CoV-2 multisystemic pathogenesis based on the organ involved that highlighted the need for considering organ-specific treatments or even personalized therapy. Besides, some FDA-approved drugs were proposed as the potential therapeutic candidates for each infected cell line.

De novo design of novel protease inhibitor candidates in the treatment of SARS-CoV-2 using deep learning, docking, and molecular dynamic simulations.

The main protease of SARS-CoV-2 is a critical target for the design and development of antiviral drugs. 2.5 M compounds were used in this study to train an LSTM generative network via transfer learning in order to identify the four best candidates capable of inhibiting the main proteases in SARS-CoV-2. The network was fine-tuned over ten generations, with each generation resulting in higher binding affinity scores. The binding affinities and interactions between the selected candidates and the SARS-CoV-2 main protease are predicted using a molecular docking simulation using AutoDock Vina. The compounds selected have a strong interaction with the key MET 165 and Cys145 residues. Molecular dynamics (MD) simulations were run for 150ns to validate the docking results on the top four ligands. Additionally, root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), and hydrogen bond analysis strongly support these findings. Furthermore, the MM-PBSA free energy calculations revealed that these chemical molecules have stable and favorable energies, resulting in a strong binding with Mpro's binding site. This study's extensive computational and statistical analyses indicate that the selected candidates may be used as potential inhibitors against the SARS-CoV-2 in-silico environment. However, additional in-vitro, in-vivo, and clinical trials are required to demonstrate their true efficacy.

A large margin piecewise linear classifier with fusion of deep features in the diagnosis of COVID-19.

The world has experienced epidemics of coronavirus infections several times over the last two decades. Recent studies have shown that using medical imaging techniques can be useful in developing an automatic computer-aided diagnosis system to detect pandemic diseases with high accuracy at an early stage. In this study, a large margin piecewise linear classifier was developed to diagnose COVID-19 compared to a wide range of viral pneumonia, including SARS and MERS, using chest x-ray images. In the proposed method, a preprocessing pipeline was employed. Moreover, deep pre- and post-rectified linear unit (ReLU) features were extracted using the well-known VGG-Net19, which was fine-tuned to optimize transfer learning. Afterward, the canonical correlation analysis was performed for feature fusion, and fused deep features were passed into the LMPL classifier. The introduced method reached the highest performance in comparison with related state-of-the-art methods for two different schemes (normal, COVID-19, and typical viral pneumonia) and (COVID-19, SARS, and MERS pneumonia) with 99.39% and 98.86% classification accuracy, respectively.

A connectivity map-based drug repurposing study and integrative analysis of transcriptomic profiling of SARS-CoV-2 infection.

AIMS: The recent outbreak of COVID-19 has become a global health concern. There are currently no effective treatment strategies and vaccines for the treatment or prevention of this fatal disease. The current study aims to determine promising treatment options for the COVID-19 through a computational drug repurposing approach. MATERIALS AND METHODS: In this study, we focus on differentially expressed genes (DEGs), detected in SARS-CoV-2 infected cell lines including "the primary human lung epithelial cell line NHBE" and "the transformed lung alveolar cell line A549". Next, the identified DEGs are used in the connectivity map (CMap) analysis to identify similarly acting therapeutic candidates. Furthermore, to interpret lists of DEGs, pathway enrichment and protein network analysis are performed. Genes are categorized into easily interpretable pathways based on their biological functions, and overrepresentation of each pathway is tested in comparison to what is expected randomly. KEY FINDINGS: The results suggest the effectiveness of lansoprazole, folic acid, sulfamonomethoxine, tolnaftate, diclofenamide, halcinonide, saquinavir, metronidazole, ebselen, lidocaine and benzocaine, histone deacetylase (HDAC) inhibitors, heat shock protein 90 (HSP90) inhibitors, and many other clinically approved drugs as potent drugs against COVID-19 outbreak. SIGNIFICANCE: Making new drugs remain a lengthy process, so the drug repurposing approach provides an insight into the therapeutics that might be helpful in this pandemic. In this study, pathway enrichment and protein network analysis are also performed, and the effectiveness of some drugs obtained from the CMap analysis has been investigated according to previous researches.

Duration of Vascular Access Usage and Patient Survival in the First Year of Hemodialysis.

INTRODUCTION: The central venous catheter (CVC) has been shown to increase mortality in hemodialysis (HD) patients compared with the arteriovenous fistula (AVF). However, no study has examined the mortality of HD patients based on the time of conversion from the CVC to AVF. In this study, we investigated the association between patients' survival and length of time of using each access. METHODS: The C5.0 algorithm was used to find rules about the relationship between duration of the different access usage and survival. The cox model was applied to assess the association of the obtained duration categories and mortality. RESULTS: From 2367 adult patients who received maintenance HD from 2012 to 2014, 705 patients were eligible for the study. Using an AVF for more than 8 months and a CVC for less than 4.2 months had the highest one-year survival rate (91.8% and 87.4%). The hazard ratio (HR) for mortality of less than 2.8 months of AVF usage compared to the longest usage was 6.90 (95% CI: 4.60 - 10.30) before adjustment and 5.03 (95% CI: 3.20 - 8.00) after adjustment for all confounders. For the CVC, the ratio was 8.8 (95% CI: 6.00 - 13.00) when comparing more than 9.2 months of usage with the lowest usage duration before an adjustment and 6.00 (95% CI: 3.80 - 9.41) after adjustment. CONCLUSION: Our results presented that regardless of the type of initial vascular access, limiting the length of the time using CVC as well as switching to AVF could significantly improve the survival of HD patients.

Identifying Factors That Affect Patient Survival After Orthotopic Liver Transplant Using Machine-Learning Techniques.

OBJECTIVES: Survival after liver transplant depends on pretransplant, peritransplant, and posttransplant factors. Identifying effective factors for patient survival after transplant can help transplant centers make better decisions. MATERIALS AND METHODS: Our study included 902 adults who received livers from deceased donors from March 2011 to March 2014 at the Shiraz Organ Transplant Center (Shiraz, Iran). In a 3-step feature selection method, effective features of 6-month survival were extracted by (1) F statistics, Pearson chi-square, and likelihood ratio chi-square and by (2) 5 machine-learning techniques. To evaluate the performance of the machine-learning techniques, Cox regression was applied to the data set. Evaluations were based on the area under the receiver operating characteristic curve and sensitivity of models. (3) We also constructed a model using all factors identified in the previous step. RESULTS: The model predicted survival based on 26 identified effective factors. In the following order, graft failure, Aspergillus infection, acute renal failure and vascular complications after transplant, as well as graft failure diagnosis interval, previous diabetes mellitus, Model for End-Stage Liver Disease score, donor inotropic support, units of packed cell received, and previous recipient dialysis, were found to be predictive factors in patient survival. The area under the receiver operating characteristic curve and model sensitivity were 0.90 and 0.81, respectively. CONCLUSIONS: Data mining analyses can help identify effective features of patient survival after liver transplant and build models with equal or higher performance than Cox regression. The order of influential factors identified with the machine-learning model was close to clinical experiments.

Identifying mutation positions in all segments of influenza genome enables better differentiation between pandemic and seasonal strains.

Influenza has a negative sense, single-stranded, and segmented RNA. In the context of pandemic influenza research, most studies have focused on variations in the surface proteins (Hemagglutinin and Neuraminidase). However, new findings suggest that all internal and external proteins of influenza viruses can contribute in pandemic emergence, pathogenicity and increasing host range. The occurrence of the 2009 influenza pandemic and the availability of many external and internal segments of pandemic and non-pandemic sequences offer a unique opportunity to evaluate the performance of machine learning models in discrimination of pandemic from seasonal sequences using mutation positions in all segments. In this study, we hypothesized that identifying mutation positions in all segments (proteins) encoded by the influenza genome would enable pandemic and seasonal strains to be more reliably distinguished. In a large scale study, we applied a range of data mining techniques to all segments of influenza for rule discovery and discrimination of pandemic from seasonal strains. CBA (classification based on association rule mining), Ripper and Decision tree algorithms were utilized to extract association rules among mutations. CBA outperformed the other models. Our approach could discriminate pandemic sequences from seasonal ones with more than 95% accuracy for PA and NP, 99.33% accuracy for NA and 100% accuracy, precision, specificity and sensitivity (recall) for M1, M2, PB1, NS1, and NS2. The values of precision, specificity, and sensitivity were more than 90% for other segments except PB2. If sequences of all segments of one strain were available, the accuracy of discrimination of pandemic strains was 100%. General rules extracted by rule base classification approaches, such as M1-V147I, NP-N334H, NS1-V112I, and PB1-L364I, were able to detect pandemic sequences with high accuracy. We observed that mutations on internal proteins of influenza can contribute in distinguishing the pandemic viruses, similar to the external ones.

Multi-source dataset for urban computing in a Smart City.

It is vital to capture and analyze, from various sources in smart cities, the data that are beneficial in urban planning and decision making for governments and individuals. Urban policy makers can find a suitable solution for urban development by using the opportunities and capacities of big data, and by combining different heterogeneous data resources in smart cities. This paper presents data related to urban computing with an aim of assessing the knowledge that can be obtained through integration of multiple independent data sources in Smart Cities. The data contains multiple sources in the city of Aarhus, Denmark from August 1, 2014 to September 30, 2014. The sources include land use, waterways, water barriers, buildings, roads, amenities, POI, weather, traffic, pollution, and parking lot data. The published data in this paper is an extended version of the City Pulse project data to which additional data sources collected from online sources have been added.

Novel approach for identification of influenza virus host range and zoonotic transmissible sequences by determination of host-related associative positions in viral genome segments.

BACKGROUND: Recent (2013 and 2009) zoonotic transmission of avian or porcine influenza to humans highlights an increase in host range by evading species barriers. Gene reassortment or antigenic shift between viruses from two or more hosts can generate a new life-threatening virus when the new shuffled virus is no longer recognized by antibodies existing within human populations. There is no large scale study to help understand the underlying mechanisms of host transmission. Furthermore, there is no clear understanding of how different segments of the influenza genome contribute in the final determination of host range. METHODS: To obtain insight into the rules underpinning host range determination, various supervised machine learning algorithms were employed to mine reassortment changes in different viral segments in a range of hosts. Our multi-host dataset contained whole segments of 674 influenza strains organized into three host categories: avian, human, and swine. Some of the sequences were assigned to multiple hosts. In point of fact, the datasets are a form of multi-labeled dataset and we utilized a multi-label learning method to identify discriminative sequence sites. Then algorithms such as CBA, Ripper, and decision tree were applied to extract informative and descriptive association rules for each viral protein segment. RESULT: We found informative rules in all segments that are common within the same host class but varied between different hosts. For example, for infection of an avian host, HA14V and NS1230S were the most important discriminative and combinatorial positions. CONCLUSION: Host range identification is facilitated by high support combined rules in this study. Our major goal was to detect discriminative genomic positions that were able to identify multi host viruses, because such viruses are likely to cause pandemic or disastrous epidemics.

A Multiple-Classifier Framework for Parkinson's Disease Detection Based on Various Vocal Tests.

Recently, speech pattern analysis applications in building predictive telediagnosis and telemonitoring models for diagnosing Parkinson's disease (PD) have attracted many researchers. For this purpose, several datasets of voice samples exist; the UCI dataset named "Parkinson Speech Dataset with Multiple Types of Sound Recordings" has a variety of vocal tests, which include sustained vowels, words, numbers, and short sentences compiled from a set of speaking exercises for healthy and people with Parkinson's disease (PWP). Some researchers claim that summarizing the multiple recordings of each subject with the central tendency and dispersion metrics is an efficient strategy in building a predictive model for PD. However, they have overlooked the point that a PD patient may show more difficulty in pronouncing certain terms than the other terms. Thus, summarizing the vocal tests may lead into loss of valuable information. In order to address this issue, the classification setting must take what has been said into account. As a solution, we introduced a new framework that applies an independent classifier for each vocal test. The final classification result would be a majority vote from all of the classifiers. When our methodology comes with filter-based feature selection, it enhances classification accuracy up to 15%.

Knowledge discovery and sequence-based prediction of pandemic influenza using an integrated classification and association rule mining (CBA) algorithm.

Pandemic influenza is a major concern worldwide. Availability of advanced technologies and the nucleotide sequences of a large number of pandemic and non-pandemic influenza viruses in 2009 provide a great opportunity to investigate the underlying rules of pandemic induction through data mining tools. Here, for the first time, an integrated classification and association rule mining algorithm (CBA) was used to discover the rules underpinning alteration of non-pandemic sequences to pandemic ones. We hypothesized that the extracted rules can lead to the development of an efficient expert system for prediction of influenza pandemics. To this end, we used a large dataset containing 5373 HA (hemagglutinin) segments of the 2009 H1N1 pandemic and non-pandemic influenza sequences. The analysis was carried out for both nucleotide and protein sequences. We found a number of new rules which potentially present the undiscovered antigenic sites at influenza structure. At the nucleotide level, alteration of thymine (T) at position 260 was the key discriminating feature in distinguishing non-pandemic from pandemic sequences. At the protein level, rules including I233K, M334L were the differentiating features. CBA efficiently classifies pandemic and non-pandemic sequences with high accuracy at both the nucleotide and protein level. Finding hotspots in influenza sequences is a significant finding as they represent the regions with low antibody reactivity. We argue that the virus breaks host immunity response by mutation at these spots. Based on the discovered rules, we developed the software, "Prediction of Pandemic Influenza" for discrimination of pandemic from non-pandemic sequences. This study opens a new vista in discovery of association rules between mutation points during evolution of pandemic influenza.

Educational level and age as contributing factors to road traffic accidents.

OBJECTIVE: This research analyzes data on road traffic accidents (RTA) in Fars province, whose roads are among the highly dangerous ones in Iran. It investigates educational level and age involved in RTA in order to discover patterns that can prevent or decrease accidents. METHODS: This research made use of data visualization techniques to find hidden patterns. The data included mortality rate related to RTA in Fars province and were obtained from Fars Forensic Medicine Registry covering a period of 1 year from March 21, 2010 to March 21, 2011. All data were analyzed using SPSS 11.5. The results were reported as descriptive indices such as frequency (percentage). The Chi-square test was applied to the data concerning educational level and age. P value less than 0.05 was considered significant. RESULTS: In the mentioned period, 1 831 people were killed, out of whom un/lowly educated people (69.6%) accounted for the highest mortality rate. The significant relationship between educational level and mortality rate was found (X2 equal to 275.98, P less than 0.0001).Also three was a significant association between age and mortality rate (X2 equal to 371.20, P less than 0.0001). Young people (age between 20 and 29 years) contribute to higher RTA mortality rate compared with other age groups. CONCLUSION: The educational level and age are significantly correlated to mortality rate. The youth and un/lowly educated people suffer more fatal RTA.

Association between physical activity and urinary incontinence in a community-based elderly population aged 70 years and over.

OBJECTIVES: The objective of the present study was to evaluate the association between physical activity (PA) levels and urinary incontinence (UI) in a community-based elderly population aged > or =70 yr. METHODS: This population-based cross-sectional survey was conducted in 2003 using an extensive health interview for each participant. A self-reported single-item questionnaire was used to estimate different levels of PA in each subject. The prevalence of UI was estimated by the self-administered International Consultation on Incontinence Questionnaire. The study population included 676 Japanese men and women. RESULTS: The prevalence of UI was 25% (34% in women and 16% in men). After adjustment for potential confounding factors, the odds ratio (95% confidence interval) of UI compared with the lowest PA group was 0.71 (0.47-1.09) and 0.58 (0.35-0.96) in subjects exhibiting middle and high levels of PA, respectively (p for trend = 0.02). CONCLUSIONS: High PA level was independently related to a lower self-reported prevalence of UI in a community-dwelling elderly population aged > or =70 yr. Although this cross-sectional study cannot demonstrate a temporal relationship between PA and the onset of UI, the findings suggest that PA may have a potentially beneficial effect on the prevention of UI. A prospective study or randomized trials are required to clarify the causality.