Search

Objective The purpose of this study was to develop a natural language processing (NLP) pipeline to identify incidental thyroid nodules (ITNs) meeting criteria for sonographic follow-up and to assess both adherence rates to white paper recommendations and downstream outcomes related to these incidental findings. Methods 21583 non-contrast chest CT reports from 2017 and 2018 were retrospectively evaluated to identify reports which included either an explicit recommendation for thyroid ultrasound, a description of a nodule ≥ 1.5 cm, or description of a nodule with suspicious features. Reports from 2018 were used to train an NLP algorithm called fastText for automated identification of such reports. Algorithm performance was then evaluated on the 2017 reports. Next, any patient from 2017 with a report meeting criteria for ultrasound follow-up was further evaluated with manual chart review to determine follow-up adherence rates and nodule-related outcomes. Results NLP identified reports with ITNs meeting criteria for sonographic follow-up with an accuracy of 96.5% (95% CI 96.2-96.7) and sensitivity of 92.1% (95% CI 89.8-94.3). In 10006 chest CTs from 2017, ITN follow-up ultrasound was indicated according to white paper criteria in 81 patients (0.8%), explicitly recommended in 46.9% (38/81) of patients, and obtained in less than half of patients in which it was appropriately recommended (17/35, 48.6%). Discussion NLP accurately identified chest CT reports meeting criteria for ITN ultrasound follow-up. Radiologist adherence to white paper guidelines and subsequent referrer adherence to radiologist recommendations showed room for improvement.

Function-as-a-Service (FaaS) has raised a growing interest in how to "tame" serverless to enable domain-specific use cases such as data-intensive applications and machine learning (ML), to name a few. Recently, several systems have been implemented for training ML models. Certainly, these research articles are significant steps in the correct direction. However, they do not completely answer the nagging question of when serverless ML training can be more cost-effective compared to traditional "serverful" computing. To help in this task, we propose MLLess, a FaaS-based ML training prototype built atop IBM Cloud Functions. To boost cost-efficiency, MLLess implements two key optimizations: a significance filter and a scale-in auto-tuner, and leverages them to specialize model training to the FaaS model. Our results certify that MLLess can be 15X faster than serverful ML systems [24] at a lower cost for ML models (such as sparse logistic regression and matrix factorization) that exhibit fast convergence.

Over 100 fatalities and more than 8000 injuries are reported on average every day in the US caused by motor vehicle accidents. In order to provide drivers a safer travel plan, we present a machine learning powered risk profiler for road segments using geo-spatial data. We built an end-to-end pipeline to extract static road features from map data and combined them with other data such as weather and traffic patterns. Our approach proposes novel methods for data pre-processing and feature engineering using statistical and clustering methods. Our model achieves significant performance improvement for risk prediction using hyper-parameter optimization (HPO) and the open source AutoGluon library to optimize the ML model. Finally, an enduser visualization interface is developed in the form of interactive maps. The results indicate 31% improvement in model performance compared to baseline when model is applied to a new geo location. We tested this approach on six major cities in the US. The findings of this research will provide users a tool to quantitatively assess accident risk at road segment level.

Motivation Drug repositioning is an attractive alternative to de novo drug discovery due to reduced time and costs to bring drugs to market. Computational repositioning methods, particularly non-black-box methods that can account for and predict a drug’s mechanism, may provide great benefit for directing future development. By tuning both data and algorithm to utilize relationships important to drug mechanisms, a computational repositioning algorithm can be trained to both predict and explain mechanistically novel indications. Results In this work, we examined the 123 curated drug mechanism paths found in the drug mechanism database (DrugMechDB) and after identifying the most important relationships, we integrated 18 data sources to produce a heterogeneous knowledge graph, MechRepoNet, capable of capturing the information in these paths. We applied the Rephetio repurposing algorithm to MechRepoNet using only a subset of relationships known to be mechanistic in nature and found adequate predictive ability on an evaluation set with AUROC value of 0.83. The resulting repurposing model allowed us to prioritize paths in our knowledge graph to produce a predicted treatment mechanism. We found that DrugMechDB paths, when present in the network were rated highly among predicted mechanisms. We then demonstrated MechRepoNet’s ability to use mechanistic insight to identify a drug’s mechanistic target, with a mean reciprocal rank of 0.525 on a test set of known drug–target interactions. Finally, we walked through repurposing examples of the anti-cancer drug imatinib for use in the treatment of asthma, and metolazone for use in the treatment of osteoporosis, to demonstrate this method’s utility in providing mechanistic insight into repurposing predictions it provides. Availability and implementation The Python code to reproduce the entirety of this analysis is available at: https://github.com/SuLab/MechRepoNet (archived at https://doi.org/10.5281/zenodo.6456335). Supplementary information Supplementary data are available at Bioinformatics online.

Motivation A phylogenetic network is a powerful model to represent entangled evolutionary histories with both divergent (speciation) and convergent (e.g. hybridization, reassortment, recombination) evolution. The standard approach to inference of hybridization networks is to (i) reconstruct rooted gene trees and (ii) leverage gene tree discordance for network inference. Recently, we introduced a method called RF-Net for accurate inference of virus reassortment and hybridization networks from input gene trees in the presence of errors commonly found in phylogenetic trees. While RF-Net demonstrated the ability to accurately infer networks with up to four reticulations from erroneous input gene trees, its application was limited by the number of reticulations it could handle in a reasonable amount of time. This limitation is particularly restrictive in the inference of the evolutionary history of segmented RNA viruses such as influenza A virus (IAV), where reassortment is one of the major mechanisms shaping the evolution of these pathogens. Results Here, we expand the functionality of RF-Net that makes it significantly more applicable in practice. Crucially, we introduce a fast extension to RF-Net, called Fast-RF-Net, that can handle large numbers of reticulations without sacrificing accuracy. In addition, we develop automatic stopping criteria to select the appropriate number of reticulations heuristically and implement a feature for RF-Net to output error-corrected input gene trees. We then conduct a comprehensive study of the original method and its novel extensions and confirm their efficacy in practice using extensive simulation and empirical IAV evolutionary analyses. Availability and implementation RF-Net 2 is available at https://github.com/flu-crew/rf-net-2. Supplementary information Supplementary data are available at Bioinformatics online.

Motivation In the last decade, de novo protein structure prediction accuracy for individual proteins has improved significantly by utilising deep learning (DL) methods for harvesting the co-evolution information from large multiple sequence alignments (MSAs). The same approach can, in principle, also be used to extract information about evolutionary-based contacts across protein–protein interfaces. However, most earlier studies have not used the latest DL methods for inter-chain contact distance prediction. This article introduces a fold-and-dock method based on predicted residue-residue distances with trRosetta. Results The method can simultaneously predict the tertiary and quaternary structure of a protein pair, even when the structures of the monomers are not known. The straightforward application of this method to a standard dataset for protein–protein docking yielded limited success. However, using alternative methods for generating MSAs allowed us to dock accurately significantly more proteins. We also introduced a novel scoring function, PconsDock, that accurately separates 98% of correctly and incorrectly folded and docked proteins. The average performance of the method is comparable to the use of traditional, template-based or ab initio shape-complementarity-only docking methods. Moreover, the results of conventional and fold-and-dock approaches are complementary, and thus a combined docking pipeline could increase overall docking success significantly. This methodology contributed to the best model for one of the CASP14 oligomeric targets, H1065. Availability and implementation All scripts for predictions and analysis are available from https://github.com/ElofssonLab/bioinfo-toolbox/ and https://gitlab.com/ElofssonLab/benchmark5/. All models joined alignments, and evaluation results are available from the following figshare repository https://doi.org/10.6084/m9.figshare.14654886.v2. Supplementary information Supplementary data are available at Bioinformatics online.

Methods to quantify biological aging are emerging as new measurement tools for epidemiology and population science and have been proposed as surrogate measures for healthy lifespan extension in geroscience clinical trials. Publicly available software packages to compute biological aging measurements from DNA methylation data have accelerated dissemination of these measures and generated rapid gains in knowledge about how different measures perform in a range of datasets. Biological age measures derived from blood chemistry data were introduced at the same time as the DNA methylation measures and, in multiple studies, demonstrate superior performance to these measures in prediction of healthy lifespan. However, their dissemination has been slow by comparison, resulting in a significant gap in knowledge. We developed a software package to help address this knowledge gap. The BioAge R package, available for download at GitHub (http://github.com/dayoonkwon/BioAge), implements three published methods to quantify biological aging based on analysis of chronological age and mortality risk: Klemera-Doubal biological age, PhenoAge, and homeostatic dysregulation. The package allows users to parametrize measurement algorithms using custom sets of biomarkers, to compare the resulting measurements to published versions of the Klemera-Doubal method and PhenoAge algorithms, and to score the measurements in new datasets. We applied BioAge to safety lab data from the CALERIE™ randomized controlled trial, the first-ever human trial of long-term calorie restriction in healthy, non-obese adults, to test effects of intervention on biological aging. Results contribute evidence that CALERIE intervention slowed biological aging. BioAge is a toolkit to facilitate measurement of biological age for geroscience. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s11357-021-00480-5.

Background Due to the COVID-19 pandemic, the demand for remote electrocardiogram (ECG) monitoring has increased drastically in an attempt to prevent the spread of the virus and keep vulnerable individuals with less severe cases out of hospitals. Enabling clinicians to set up remote patient ECG monitoring easily and determining how to classify the ECG signals accurately so relevant alerts are sent in a timely fashion is an urgent problem to be addressed for remote patient monitoring (RPM) to be adopted widely. Hence, a new technique is required to enable routine and widespread use of RPM, as is needed due to COVID-19. Objective The primary aim of this research is to create a robust and easy-to-use solution for personalized ECG monitoring in real-world settings that is precise, easily configurable, and understandable by clinicians. Methods In this paper, we propose a Personalized Monitoring Model (PMM) for ECG data based on motif discovery. Motif discovery finds meaningful or frequently recurring patterns in patient ECG readings. The main strategy is to use motif discovery to extract a small sample of personalized motifs for each individual patient and then use these motifs to predict abnormalities in real-time readings of that patient using an artificial logical network configured by a physician. Results Our approach was tested on 30 minutes of ECG readings from 32 patients. The average diagnostic accuracy of the PMM was always above 90% and reached 100% for some parameters, compared to 80% accuracy for the Generalized Monitoring Models (GMM). Regardless of parameter settings, PMM training models were generated within 3-4 minutes, compared to 1 hour (or longer, with increasing amounts of training data) for the GMM. Conclusions Our proposed PMM almost eliminates many of the training and small sample issues associated with GMMs. It also addresses accuracy and computational cost issues of the GMM, caused by the uniqueness of heartbeats and training issues. In addition, it addresses the fact that doctors and nurses typically do not have data science training and the skills needed to configure, understand, and even trust existing black box machine learning models.

In the present study, iodide-mediated 3,3′,5,5′-tetramethylbenzidine (TMB)-H2O2 reaction system was applied to a microfluidic paper-based analytical device (μPAD) for non-enzymatic colorimetric determination of H2O2. The proposed system is portable and incorporates a μPAD with a machine learning-based smartphone app. A smartphone app called “Hi-perox Sens” capable of image capture, cropping and processing was developed to make the system simple and user-friendly. Briefly, circular μPADs were designed and tested with varying concentrations of H2O2. Following the color change, the images of the μPADs were taken with four different smartphones under seven different illumination conditions. In order to make the system more robust and adaptive against illumination variation and camera optics, the images were first processed for feature extraction and then used to train machine learning classifiers. According to the results, TMB + KI showed the highest classification accuracy (97.8%) with inter-phone repeatability at t = 30 s under versatile illumination and maintained its accuracy for 10 minutes. In addition, the performance of the system was also comparable to two different commercially available H2O2 kits in real samples.

Background: Continuous positive airway pressure (CPAP) is an effective treatment for obstructive sleep apnea (OSA), but treatment compliance is often unsatisfactory. Objective: The aim of this study was to assess the effectiveness and cost-effectiveness of an intelligent monitoring system for improving CPAP compliance. Methods: This is a prospective, open label, parallel, randomized controlled trial including 60 newly diagnosed patients with OSA requiring CPAP (Apnea–Hypopnea Index [AHI] >15) from Lleida, Spain. Participants were randomized (1:1) to standard management or the MiSAOS intelligent monitoring system, involving (1) early compliance detection, thus providing measures of patient’s CPAP compliance from the very first days of usage; (2) machine learning–based prediction of midterm future CPAP compliance; and (3) rule-based recommendations for the patient (app) and care team. Clinical and anthropometric variables, daytime sleepiness, and quality of life were recorded at baseline and after 6 months, together with patient’s compliance, satisfaction, and health care costs. Results: Randomized patients had a mean age of 57 (SD 11) years, mean AHI of 50 (SD 27), and 13% (8/60) were women. Patients in the intervention arm had a mean (95% CI) of 1.14 (0.04-2.23) hours/day higher adjusted CPAP compliance than controls (P=.047). Patients’ satisfaction was excellent in both arms, and up to 88% (15/17) of intervention patients reported willingness to keep using the MiSAOS app in the future. No significant differences were found in costs (control: mean €90.2 (SD 53.14) (US $105.76 [SD 62.31]); intervention: mean €96.2 (SD 62.13) (US $112.70 [SD 72.85]); P=.70; €1=US $1.17 was considered throughout). Overall costs combined with results on compliance demonstrated cost-effectiveness in a bootstrap-based simulation analysis. Conclusions: A machine learning–based intelligent monitoring system increased daily compliance, reported excellent patient satisfaction similar to that reported in usual care, and did not incur in a substantial increase in costs, thus proving cost-effectiveness. This study supports the implementation of intelligent eHealth frameworks for the management of patients with CPAP-treated OSA and confirms the value of patients’ empowerment in the management of chronic diseases. This work is part of the myOSA project (RTC-2014-3138-1), funded by the Spanish Ministry of Economy, Industry and Competitiveness (Ministerio de Economía, Industria y Competitividad) and Agencia Estatal de Investigación, under the framework “Retos-Colaboración”, State Scientific and Technical Research and Innovation Plan 2013-2016. The work was cofunded by the European Regional Development Fund (ERDF), “A way to make Europe”. JdB acknowledges receiving financial support from the Catalan Health Department (Pla Estratègic de Recerca i Innovació en Salut [PERIS] 2016: SLT002/16/00364) and Instituto de Salud Carlos III (ISCIII; Miguel Servet 2019: CP19/00108), co-funded by the European Social Fund (ESF), “Investing in your future”. Funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Single amino acid variations (SAVs) are a primary contributor to variations in the human genome. Identifying pathogenic SAVs can aid in the diagnosis and understanding of the genetic architecture of complex diseases, such as cancer. Most approaches for predicting the functional effects or pathogenicity of SAVs rely on either sequence or structural information. Nevertheless, previous analyses have shown that methods that depend on only sequence or structural information may have limited accuracy. Recently, researchers have attempted to increase the accuracy of their predictions by incorporating protein dynamics into pathogenicity predictions. This study presents < Lai Yang Rubenstein Uzun Sarkar > (LYRUS), a machine learning method that uses an XGBoost classifier selected by TPOT to predict the pathogenicity of SAVs. LYRUS incorporates five sequence-based features, six structure-based features, and four dynamics-based features. Uniquely, LYRUS includes a newly-proposed sequence co-evolution feature called variation number. LYRUS’s performance was evaluated using a dataset that contains 4,363 protein structures corresponding to 20,307 SAVs based on human genetic variant data from the ClinVar database. Based on our dataset, the LYRUS classifier has a higher accuracy, specificity, F-measure, and Matthews correlation coefficient (MCC) than alternative methods including PolyPhen2, PROVEAN, SIFT, Rhapsody, EVMutation, MutationAssessor, SuSPect, FATHMM, and MVP. Variation numbers used within LYRUS differ greatly between pathogenic and neutral SAVs, and have a high feature weight in the XGBoost classifier employed by this method. Applications of the method to PTEN and TP53 further corroborate LYRUS’s strong performance. LYRUS is freely available and the source code can be found at https://github.com/jiaying2508/LYRUS.

Background: Heterogeneity in Acute Respiratory Distress Syndrome (ARDS), as a consequence of its non-specific definition, has led to a multitude of negative randomised controlled trials (RCTs). Investigators have sought to identify heterogeneity of treatment effect (HTE) in RCTs using clustering algorithms. We evaluated the proficiency of several commonly-used machine-learning algorithms to identify clusters where HTE may be detected. Methods: Five unsupervised: Latent class analysis (LCA), K-means, partition around medoids, hierarchical, and spectral clustering; and four supervised algorithms: model-based recursive partitioning, Causal Forest (CF), and X-learner with Random Forest (XL-RF) and Bayesian Additive Regression Trees were individually applied to three prior ARDS RCTs. Clinical data and research protein biomarkers were used as partitioning variables, with the latter excluded for secondary analyses. For a clustering schema, HTE was evaluated based on the interaction term of treatment group and cluster with day-90 mortality as the dependent variable. Findings: No single algorithm identified clusters with significant HTE in all three trials. LCA, XL-RF, and CF identified HTE most frequently (2/3 RCTs). Important partitioning variables in the unsupervised approaches were consistent across algorithms and RCTs. In supervised models, important partitioning variables varied between algorithms and across RCTs. In algorithms where clusters demonstrated HTE in the same trial, patients frequently interchanged clusters from treatment-benefit to treatment-harm clusters across algorithms. LCA aside, results from all other algorithms were subject to significant alteration in cluster composition and HTE with random seed change. Removing research biomarkers as partitioning variables greatly reduced the chances of detecting HTE across all algorithms. Interpretation: Machine-learning algorithms were inconsistent in their abilities to identify clusters with significant HTE. Protein biomarkers were essential in identifying clusters with HTE. Investigations using machine-learning approaches to identify clusters to seek HTE require cautious interpretation. Funding: NIGMS R35 GM142992 (PS), NHLBI R35 HL140026 (CSC); NIGMS R01 GM123193, Department of Defense W81XWH-21-1-0009, NIA R21 AG068720, NIDA R01 DA051464 (MMC)

Motivation With the advancement of technology, we can generate and access large-scale, high dimensional and diverse genomics data, especially through single-cell RNA sequencing (scRNA-seq). However, integrative downstream analysis from multiple scRNA-seq datasets remains challenging due to batch effects. Results In this article, we propose a light-structured deep learning framework called ResPAN for scRNA-seq data integration. ResPAN is based on Wasserstein Generative Adversarial Network (WGAN) combined with random walk mutual nearest neighbor pairing and fully skip-connected autoencoders to reduce the differences among batches. We also discuss the limitations of existing methods and demonstrate the advantages of our model over seven other methods through extensive benchmarking studies on both simulated data under various scenarios and real datasets across different scales. Our model achieves leading performance on both batch correction and biological information conservation and maintains scalable to datasets with over half a million cells. Availability and implementation An open-source implementation of ResPAN and scripts to reproduce the results can be downloaded from: https://github.com/AprilYuge/ResPAN. Supplementary information Supplementary data are available at Bioinformatics online.

Background Online media play an important role in public health emergencies and serve as essential communication platforms. Infoveillance of online media during the COVID-19 pandemic is an important step toward gaining a better understanding of crisis communication. Objective The goal of this study was to perform a longitudinal analysis of the COVID-19–related content on online media based on natural language processing. Methods We collected a data set of news articles published by Croatian online media during the first 13 months of the pandemic. First, we tested the correlations between the number of articles and the number of new daily COVID-19 cases. Second, we analyzed the content by extracting the most frequent terms and applied the Jaccard similarity coefficient. Third, we compared the occurrence of the pandemic-related terms during the two waves of the pandemic. Finally, we applied named entity recognition to extract the most frequent entities and tracked the dynamics of changes during the observation period. Results The results showed no significant correlation between the number of articles and the number of new daily COVID-19 cases. Furthermore, there were high overlaps in the terminology used in all articles published during the pandemic with a slight shift in the pandemic-related terms between the first and the second waves. Finally, the findings indicate that the most influential entities have lower overlaps for the identified people and higher overlaps for locations and institutions. Conclusions Our study shows that online media have a prompt response to the pandemic with a large number of COVID-19–related articles. There was a high overlap in the frequently used terms across the first 13 months, which may indicate the narrow focus of reporting in certain periods. However, the pandemic-related terminology is well-covered.

Background Machine learning applications in the health care domain can have a great impact on people’s lives. At the same time, medical data is usually big, requiring a significant number of computational resources. Although this might not be a problem for the wide adoption of machine learning tools in high-income countries, the availability of computational resources can be limited in low-income countries and on mobile devices. This can limit many people from benefiting from the advancement in machine learning applications in the field of health care. Objective In this study, we explore three methods to increase the computational efficiency and reduce model sizes of either recurrent neural networks (RNNs) or feedforward deep neural networks (DNNs) without compromising their accuracy. Methods We used inpatient mortality prediction as our case analysis upon review of an intensive care unit dataset. We reduced the size of RNN and DNN by applying pruning of “unused” neurons. Additionally, we modified the RNN structure by adding a hidden layer to the RNN cell but reducing the total number of recurrent layers to accomplish a reduction of the total parameters used in the network. Finally, we implemented quantization on DNN by forcing the weights to be 8 bits instead of 32 bits. Results We found that all methods increased implementation efficiency, including training speed, memory size, and inference speed, without reducing the accuracy of mortality prediction. Conclusions Our findings suggest that neural network condensation allows for the implementation of sophisticated neural network algorithms on devices with lower computational resources.

Background Stroke, a cerebrovascular disease, is one of the major causes of death. It causes significant health and financial burdens for both patients and health care systems. One of the important risk factors for stroke is health-related behavior, which is becoming an increasingly important focus of prevention. Many machine learning models have been built to predict the risk of stroke or to automatically diagnose stroke, using predictors such as lifestyle factors or radiological imaging. However, there have been no models built using data from lab tests. Objective The aim of this study was to apply computational methods using machine learning techniques to predict stroke from lab test data. Methods We used the National Health and Nutrition Examination Survey data sets with three different data selection methods (ie, without data resampling, with data imputation, and with data resampling) to develop predictive models. We used four machine learning classifiers and six performance measures to evaluate the performance of the models. Results We found that accurate and sensitive machine learning models can be created to predict stroke from lab test data. Our results show that the data resampling approach performed the best compared to the other two data selection techniques. Prediction with the random forest algorithm, which was the best algorithm tested, achieved an accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and area under the curve of 0.96, 0.97, 0.96, 0.75, 0.99, and 0.97, respectively, when all of the attributes were used. Conclusions The predictive model, built using data from lab tests, was easy to use and had high accuracy. In future studies, we aim to use data that reflect different types of stroke and to explore the data to build a prediction model for each type.

Background Artificial intelligence (AI), such as machine learning (ML), shows great promise for improving clinical decision-making in cardiac diseases by outperforming statistical-based models. However, few AI-based tools have been implemented in cardiology clinics because of the sociotechnical challenges during transitioning from algorithm development to real-world implementation. Objective This study explored how an ML-based tool for predicting ventricular tachycardia and ventricular fibrillation (VT/VF) could support clinical decision-making in the remote monitoring of patients with an implantable cardioverter defibrillator (ICD). Methods Seven experienced electrophysiologists participated in a near-live feasibility and qualitative study, which included walkthroughs of 5 blinded retrospective patient cases, use of the prediction tool, and questionnaires and interview questions. All sessions were video recorded, and sessions evaluating the prediction tool were transcribed verbatim. Data were analyzed through an inductive qualitative approach based on grounded theory. Results The prediction tool was found to have potential for supporting decision-making in ICD remote monitoring by providing reassurance, increasing confidence, acting as a second opinion, reducing information search time, and enabling delegation of decisions to nurses and technicians. However, the prediction tool did not lead to changes in clinical action and was found less useful in cases where the quality of data was poor or when VT/VF predictions were found to be irrelevant for evaluating the patient. Conclusions When transitioning from AI development to testing its feasibility for clinical implementation, we need to consider the following: expectations must be aligned with the intended use of AI; trust in the prediction tool is likely to emerge from real-world use; and AI accuracy is relational and dependent on available information and local workflows. Addressing the sociotechnical gap between the development and implementation of clinical decision-support tools based on ML in cardiac care is essential for succeeding with adoption. It is suggested to include clinical end-users, clinical contexts, and workflows throughout the overall iterative approach to design, development, and implementation.

Background Multimorbidity clinical risk scores allow clinicians to quickly assess their patients' health for decision making, often for recommendation to care management programs. However, these scores are limited by several issues: existing multimorbidity scores (1) are generally limited to one data group (eg, diagnoses, labs) and may be missing vital information, (2) are usually limited to specific demographic groups (eg, age), and (3) do not formally provide any granularity in the form of more nuanced multimorbidity risk scores to direct clinician attention. Objective Using diagnosis, lab, prescription, procedure, and demographic data from electronic health records (EHRs), we developed a physiologically diverse and generalizable set of multimorbidity risk scores. Methods Using EHR data from a nationwide cohort of patients, we developed the total health profile, a set of six integrated risk scores reflecting five distinct organ systems and overall health. We selected the occurrence of an inpatient hospital visitation over a 2-year follow-up window, attributable to specific organ systems, as our risk endpoint. Using a physician-curated set of features, we trained six machine learning models on 794,294 patients to predict the calibrated probability of the aforementioned endpoint, producing risk scores for heart, lung, neuro, kidney, and digestive functions and a sixth score for combined risk. We evaluated the scores using a held-out test cohort of 198,574 patients. Results Study patients closely matched national census averages, with a median age of 41 years, a median income of $66,829, and racial averages by zip code of 73.8% White, 5.9% Asian, and 11.9% African American. All models were well calibrated and demonstrated strong performance with areas under the receiver operating curve (AUROCs) of 0.83 for the total health score (THS), 0.89 for heart, 0.86 for lung, 0.84 for neuro, 0.90 for kidney, and 0.83 for digestive functions. There was consistent performance of this scoring system across sexes, diverse patient ages, and zip code income levels. Each model learned to generate predictions by focusing on appropriate clinically relevant patient features, such as heart-related hospitalizations and chronic hypertension diagnosis for the heart model. The THS outperformed the other commonly used multimorbidity scoring systems, specifically the Charlson Comorbidity Index (CCI) and the Elixhauser Comorbidity Index (ECI) overall (AUROCs: THS=0.823, CCI=0.735, ECI=0.649) as well as for every age, sex, and income bracket. Performance improvements were most pronounced for middle-aged and lower-income subgroups. Ablation tests using only diagnosis, prescription, social determinants of health, and lab feature groups, while retaining procedure-related features, showed that the combination of feature groups has the best predictive performance, though only marginally better than the diagnosis-only model on at-risk groups. Conclusions Massive retrospective EHR data sets have made it possible to use machine learning to build practical multimorbidity risk scores that are highly predictive, personalizable, intuitive to explain, and generalizable across diverse patient populations.

Background Although the COVID-19 pandemic has left an unprecedented impact worldwide, countries such as the United States have reported the most substantial incidence of COVID-19 cases worldwide. Within the United States, various sociodemographic factors have played a role in the creation of regional disparities. Regional disparities have resulted in the unequal spread of disease between US counties, underscoring the need for efficient and accurate predictive modeling strategies to inform public health officials and reduce the burden on health care systems. Furthermore, despite the widespread accessibility of COVID-19 vaccines across the United States, vaccination rates have become stagnant, necessitating predictive modeling to identify important factors impacting vaccination uptake. Objective This study aims to determine the association between sociodemographic factors and vaccine uptake across counties in the United States. Methods Sociodemographic data on fully vaccinated and unvaccinated individuals were sourced from several online databases such as the US Centers for Disease Control and Prevention and the US Census Bureau COVID-19 Site. Machine learning analysis was performed using XGBoost and sociodemographic data. Results Our model predicted COVID-19 vaccination uptake across US counties with 62% accuracy. In addition, it identified location, education, ethnicity, income, and household access to the internet as the most critical sociodemographic features in predicting vaccination uptake in US counties. Lastly, the model produced a choropleth demonstrating areas of low and high vaccination rates, which can be used by health care authorities in future pandemics to visualize and prioritize areas of low vaccination and design targeted vaccination campaigns. Conclusions Our study reveals that sociodemographic characteristics are predictors of vaccine uptake rates across counties in the United States and, if leveraged appropriately, can assist policy makers and public health officials to understand vaccine uptake rates and craft policies to improve them.

Background The absolute number of femoral neck fractures (FNFs) is increasing; however, the prediction of traumatic femoral head necrosis remains difficult. Machine learning algorithms have the potential to be superior to traditional prediction methods for the prediction of traumatic femoral head necrosis. Objective The aim of this study is to use machine learning to construct a model for the analysis of risk factors and prediction of osteonecrosis of the femoral head (ONFH) in patients with FNF after internal fixation. Methods We retrospectively collected preoperative, intraoperative, and postoperative clinical data of patients with FNF in 4 hospitals in Shanghai and followed up the patients for more than 2.5 years. A total of 259 patients with 43 variables were included in the study. The data were randomly divided into a training set (181/259, 69.8%) and a validation set (78/259, 30.1%). External data (n=376) were obtained from a retrospective cohort study of patients with FNF in 3 other hospitals. Least absolute shrinkage and selection operator regression and the support vector machine algorithm were used for variable selection. Logistic regression, random forest, support vector machine, and eXtreme Gradient Boosting (XGBoost) were used to develop the model on the training set. The validation set was used to tune the model hyperparameters to determine the final prediction model, and the external data were used to compare and evaluate the model performance. We compared the accuracy, discrimination, and calibration of the models to identify the best machine learning algorithm for predicting ONFH. Shapley additive explanations and local interpretable model-agnostic explanations were used to determine the interpretability of the black box model. Results A total of 11 variables were selected for the models. The XGBoost model performed best on the validation set and external data. The accuracy, sensitivity, and area under the receiver operating characteristic curve of the model on the validation set were 0.987, 0.929, and 0.992, respectively. The accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve of the model on the external data were 0.907, 0.807, 0.935, and 0.933, respectively, and the log-loss was 0.279. The calibration curve demonstrated good agreement between the predicted probability and actual risk. The interpretability of the features and individual predictions were realized using the Shapley additive explanations and local interpretable model-agnostic explanations algorithms. In addition, the XGBoost model was translated into a self-made web-based risk calculator to estimate an individual’s probability of ONFH. Conclusions Machine learning performs well in predicting ONFH after internal fixation of FNF. The 6-variable XGBoost model predicted the risk of ONFH well and had good generalization ability on the external data, which can be used for the clinical prediction of ONFH after internal fixation of FNF.

Background Cancer genetic testing to assess an individual’s cancer risk and to enable genomics-informed cancer treatment has grown exponentially in the past decade. Because of this continued growth and a shortage of health care workers, there is a need for automated strategies that provide high-quality genetics services to patients to reduce the clinical demand for genetics providers. Conversational agents have shown promise in managing mental health, pain, and other chronic conditions and are increasingly being used in cancer genetic services. However, research on how patients interact with these agents to satisfy their information needs is limited. Objective Our primary aim is to assess user interactions with a conversational agent for pretest genetics education. Methods We conducted a feasibility study of user interactions with a conversational agent who delivers pretest genetics education to primary care patients without cancer who are eligible for cancer genetic evaluation. The conversational agent provided scripted content similar to that delivered in a pretest genetic counseling visit for cancer genetic testing. Outside of a core set of information delivered to all patients, users were able to navigate within the chat to request additional content in their areas of interest. An artificial intelligence–based preprogrammed library was also established to allow users to ask open-ended questions to the conversational agent. Transcripts of the interactions were recorded. Here, we describe the information selected, time spent to complete the chat, and use of the open-ended question feature. Descriptive statistics were used for quantitative measures, and thematic analyses were used for qualitative responses. Results We invited 103 patients to participate, of which 88.3% (91/103) were offered access to the conversational agent, 39% (36/91) started the chat, and 32% (30/91) completed the chat. Most users who completed the chat indicated that they wanted to continue with genetic testing (21/30, 70%), few were unsure (9/30, 30%), and no patient declined to move forward with testing. Those who decided to test spent an average of 10 (SD 2.57) minutes on the chat, selected an average of 1.87 (SD 1.2) additional pieces of information, and generally did not ask open-ended questions. Those who were unsure spent 4 more minutes on average (mean 14.1, SD 7.41; P=.03) on the chat, selected an average of 3.67 (SD 2.9) additional pieces of information, and asked at least one open-ended question. Conclusions The pretest chat provided enough information for most patients to decide on cancer genetic testing, as indicated by the small number of open-ended questions. A subset of participants were still unsure about receiving genetic testing and may require additional education or interpersonal support before making a testing decision. Conversational agents have the potential to become a scalable alternative for pretest genetics education, reducing the clinical demand on genetics providers.

Background The COVID-19 global pandemic has increased the burden of mental illness on Canadian adults. However, the complex combination of demographic, economic, and lifestyle factors and perceived health risks contributing to patterns of anxiety and depression has not been explored. Objective The aim of this study is to harness flexible machine learning methods to identify constellations of factors related to symptoms of mental illness and to understand their changes over time during the COVID-19 pandemic. Methods Cross-sectional samples of Canadian adults (aged ≥18 years) completed web-based surveys in 6 waves from May to December 2020 (N=6021), and quota sampling strategies were used to match the English-speaking Canadian population in age, gender, and region. The surveys measured anxiety and depression symptoms, sociodemographic characteristics, substance use, and perceived COVID-19 risks and worries. First, principal component analysis was used to condense highly comorbid anxiety and depression symptoms into a single data-driven measure of emotional distress. Second, eXtreme Gradient Boosting (XGBoost), a machine learning algorithm that can model nonlinear and interactive relationships, was used to regress this measure on all included explanatory variables. Variable importance and effects across time were explored using SHapley Additive exPlanations (SHAP). Results Principal component analysis of responses to 9 anxiety and depression questions on an ordinal scale revealed a primary latent factor, termed “emotional distress,” that explained 76% of the variation in all 9 measures. Our XGBoost model explained a substantial proportion of variance in emotional distress (r2=0.39). The 3 most important items predicting elevated emotional distress were increased worries about finances (SHAP=0.17), worries about getting COVID-19 (SHAP=0.17), and younger age (SHAP=0.13). Hopefulness was associated with emotional distress and moderated the impacts of several other factors. Predicted emotional distress exhibited a nonlinear pattern over time, with the highest predicted symptoms in May and November and the lowest in June. Conclusions Our results highlight factors that may exacerbate emotional distress during the current pandemic and possible future pandemics, including a role of hopefulness in moderating distressing effects of other factors. The pandemic disproportionately affected emotional distress among younger adults and those economically impacted.

Background SMS text messages as a form of mobile health are increasingly being used to support individuals with chronic diseases in novel ways that leverage the mobility and capabilities of mobile phones. However, there are knowledge gaps in mobile health, including how to maximize engagement. Objective This study aims to categorize program SMS text messages and participant replies using machine learning (ML) and to examine whether message characteristics are associated with premature program stopping and engagement. Methods We assessed communication logs from SMS text message–based chronic disease prevention studies that encouraged 1-way (SupportMe/ITM) and 2-way (TEXTMEDS [Text Messages to Improve Medication Adherence and Secondary Prevention]) communication. Outgoing messages were manually categorized into 5 message intents (informative, instructional, motivational, supportive, and notification) and replies into 7 groups (stop, thanks, questions, reporting healthy, reporting struggle, general comment, and other). Grid search with 10-fold cross-validation was implemented to identify the best-performing ML models and evaluated using nested cross-validation. Regression models with interaction terms were used to compare the association of message intent with premature program stopping and engagement (replied at least 3 times and did not prematurely stop) in SupportMe/ITM and TEXTMEDS. Results We analyzed 1550 messages and 4071 participant replies. Approximately 5.49% (145/2642) of participants responded with stop, and 11.7% (309/2642) of participants were engaged. Our optimal ML model correctly classified program message intent with 76.6% (95% CI 63.5%-89.8%) and replies with 77.8% (95% CI 74.1%-81.4%) balanced accuracy (average area under the curve was 0.95 and 0.96, respectively). Overall, supportive (odds ratio [OR] 0.53, 95% CI 0.35-0.81) messages were associated with reduced chance of stopping, as were informative messages in SupportMe/ITM (OR 0.35, 95% CI 0.20-0.60) but not in TEXTMEDS (for interaction, P Conclusions ML models enable monitoring and detailed characterization of program messages and participant replies. Outgoing message intent may influence premature program stopping and engagement, although the strength and direction of association appear to vary by program type. Future studies will need to examine whether modifying message characteristics can optimize engagement and whether this leads to behavior change.

Background Nonvalvular atrial fibrillation (NVAF) affects almost 6 million Americans and is a major contributor to stroke but is significantly undiagnosed and undertreated despite explicit guidelines for oral anticoagulation. Objective The aim of this study is to investigate whether the use of semisupervised natural language processing (NLP) of electronic health record’s (EHR) free-text information combined with structured EHR data improves NVAF discovery and treatment and perhaps offers a method to prevent thousands of deaths and save billions of dollars. Methods We abstracted 96,681 participants from the University of Buffalo faculty practice’s EHR. NLP was used to index the notes and compare the ability to identify NVAF, congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, stroke or transient ischemic attack, vascular disease, age 65 to 74 years, sex category (CHA2DS2-VASc), and Hypertension, Abnormal liver/renal function, Stroke history, Bleeding history or predisposition, Labile INR, Elderly, Drug/alcohol usage (HAS-BLED) scores using unstructured data (International Classification of Diseases codes) versus structured and unstructured data from clinical notes. In addition, we analyzed data from 63,296,120 participants in the Optum and Truven databases to determine the NVAF frequency, rates of CHA2DS2‑VASc ≥2, and no contraindications to oral anticoagulants, rates of stroke and death in the untreated population, and first year’s costs after stroke. Results The structured-plus-unstructured method would have identified 3,976,056 additional true NVAF cases (PUS $13.5 billion. Conclusions Artificial intelligence–informed bio-surveillance combining NLP of free-text information with structured EHR data improves data completeness, prevents thousands of strokes, and saves lives and funds. This method is applicable to many disorders with profound public health consequences.

Background Assessing patient-reported outcomes (PROs) through interviews or conversations during clinical encounters provides insightful information about survivorship. Objective This study aims to test the validity of natural language processing (NLP) and machine learning (ML) algorithms in identifying different attributes of pain interference and fatigue symptoms experienced by child and adolescent survivors of cancer versus the judgment by PRO content experts as the gold standard to validate NLP/ML algorithms. Methods This cross-sectional study focused on child and adolescent survivors of cancer, aged 8 to 17 years, and caregivers, from whom 391 meaning units in the pain interference domain and 423 in the fatigue domain were generated for analyses. Data were collected from the After Completion of Therapy Clinic at St. Jude Children’s Research Hospital. Experienced pain interference and fatigue symptoms were reported through in-depth interviews. After verbatim transcription, analyzable sentences (ie, meaning units) were semantically labeled by 2 content experts for each attribute (physical, cognitive, social, or unclassified). Two NLP/ML methods were used to extract and validate the semantic features: bidirectional encoder representations from transformers (BERT) and Word2vec plus one of the ML methods, the support vector machine or extreme gradient boosting. Receiver operating characteristic and precision-recall curves were used to evaluate the accuracy and validity of the NLP/ML methods. Results Compared with Word2vec/support vector machine and Word2vec/extreme gradient boosting, BERT demonstrated higher accuracy in both symptom domains, with 0.931 (95% CI 0.905-0.957) and 0.916 (95% CI 0.887-0.941) for problems with cognitive and social attributes on pain interference, respectively, and 0.929 (95% CI 0.903-0.953) and 0.917 (95% CI 0.891-0.943) for problems with cognitive and social attributes on fatigue, respectively. In addition, BERT yielded superior areas under the receiver operating characteristic curve for cognitive attributes on pain interference and fatigue domains (0.923, 95% CI 0.879-0.997; 0.948, 95% CI 0.922-0.979) and superior areas under the precision-recall curve for cognitive attributes on pain interference and fatigue domains (0.818, 95% CI 0.735-0.917; 0.855, 95% CI 0.791-0.930). Conclusions The BERT method performed better than the other methods. As an alternative to using standard PRO surveys, collecting unstructured PROs via interviews or conversations during clinical encounters and applying NLP/ML methods can facilitate PRO assessment in child and adolescent cancer survivors.

Today there is a rapid development taking place in phenotyping of plants using non-destructive image based machine vision techniques. Machine vision based plant phenotyping ranges from single plant trait estimation to broad assessment of crop canopy for thousands of plants in the field. Plant phenotyping systems either use single imaging method or integrative approach signifying simultaneous use of some of the imaging techniques like visible red, green and blue (RGB) imaging, thermal imaging, chlorophyll fluorescence imaging (CFIM), hyperspectral imaging, 3-dimensional (3-D) imaging or high resolution volumetric imaging. This paper provides an overview of imaging techniques and their applications in the field of plant phenotyping. This paper presents a comprehensive survey on recent machine vision methods for plant trait estimation and classification. In this paper, information about publicly available datasets is provided for uniform comparison among the state-of-the-art phenotyping methods. This paper also presents future research directions related to the use of deep learning based machine vision algorithms for structural (2-D and 3-D), physiological and temporal trait estimation, and classification studies in plants.

This paper aims to predict a set of apps a user will open on her mobile device in the next time slot. Such an information is essential for many smartphone operations, e.g., app pre-loading and content pre-caching, to improve user experience. However, it is hard to build an explicit model that accurately captures the complex environment context and predicts a set of apps at one time. This paper presents a deep reinforcement learning framework, named as DeepAPP, which learns a model-free predictive neural network from historical app usage data. Meanwhile, an online updating strategy is designed to adapt the predictive network to the time-varying app usage behavior. To transform DeepAPP into a practical deep reinforcement learning system, several challenges are addressed by developing a context representation method for complex contextual environment, a general agent for overcoming data sparsity and a lightweight personalized agent for minimizing the prediction time. Extensive experiments on a large-scale anonymized app usage dataset reveal that DeepAPP provides high accuracy (precision 70.6% and recall of 62.4%) and reduces the prediction time of the state-of-the-art by 6.58 times. A field experiment of 29 participants demonstrates DeepAPP can effectively reduce launch time of apps.

Enforcing admired machine learning approaches to huge data enhanced novel issues for researchers. Conventional libraries could not suitably fulfil the requirement of complex model with wide variety of data and system parameters. Therefore new methodologies are required performing the computation on more than one machine over distributed environment. Some distributed frameworks on huge data such as MapReduce and TensorFlow have been deployed to solve various machine learning problems in heterogeneous distributed environment. The objective of this paper is providing a wide variety of useful information about platforms, approaches, problems, datasets, and optimization approaches in distributed systems. So researchers have been used the beneficial information to develop new approaches for efficient machine learning. This paper also covers the various formats of data like structured, semi structured and unstructured big data. A brief review of previous works is also represented in text and tabular format to provide motivation to the researchers for developing new paradigm of distributed computing environment.

This paper is a unique effort to resolve the scaling issues of machine learning using the multi-node environment of Big Data. For this purpose, it incorporates the concept of machine learning with big data. Machine learning is a branch of Artificial Intelligence which trains the computers to learn without being explicitly programmed. Machine learning works on the development of computer programs and software that work according to the input dataset. In machine learning, classification is used to identify the class of instances. In this the category or class or a group provides the new observation to which it belongs on the basis of the training dataset. Training datasets contain those instances whose class is already known. An algorithm that implements classification is known as a classifier. The main aim of this paper is to evaluate how different classifiers work on two different datasets, one is a real time dataset of MP Election nomination (2018), while the other is a standard Breast Cancer dataset, released by University of Wisconsin (1995). There are various classifiers available like Decision Trees, K-Nearest Neighbors, Support Vector Machine, Logistic Regression, Naive Bayes etc. Different classifiers work differently on different datasets, giving different accuracy according to the kind and the size of the dataset. So we have used different classifiers for our work. When we were applying machine learning classifier on a real-time dataset, i.e. MPSEC dataset, Accuracy goes down. Similarly, these classifiers apply to Standard Breast Cancer dataset then results show outstanding accuracy. This research work shows the exciting patterns to study the case of a real-time and standard dataset using different machine learning classifiers with a combination of big data. This research work also incorporates the concept of parallelization using a multi-nodes environment which is able to deal with scalability issues.

Malware propagation by adversaries has witnessed many issues across the globe. Often it is found that malware is released in different countries for monetary gains. With the proliferation of malware spreading activities, it is made possible that now we have malware patterns that are used for training machine learning models. Thus machine learning became indispensable for malware detection. The traditional machine learning models have limitations in performance as the training depth is limited. The emergence of deep learning models paved way for more training possibilities and improvement in detection accuracy with least false positives. This paper reviews literature on deep learning techniques that are used for malware detection. The deep learning methods used for malware detection include CNN, RNN, LSTM and auto encoders. LSTM is found to have memory in the cell to have better possibilities. Auto encoders are found to have better unsupervised approach with encoding and decoding to arrive at abnormalities (malware) detection. There are many contributions found using machine learning and deep learning towards Android malware detection. This paper provides knowledge that leads to further research in deep learning which is essential to improve the state of the art.

The concept of Cyber Physical System (CPS) is widely used in different industries across the globe. In fact, it is the holistic approach towards dealing with cyber space and physical environments that do have inter-dependencies. In the existing systems, there was a separate approach for security of the two worlds (cyber and physical). However, it does not provide necessary security when security is employed independently. The research in this paper identifies the need for integrated security for CPS. Besides it throws light into different security challenges associated with CPS and the countermeasures that existed based on machine learning and deep learning techniques that come under Artificial Intelligence (AI) and data science. From the review of literature, it is understood that data science perspective is suitable for protecting CPS with required adaptive strategy. This paper provides several useful insights related to security analysis of CPS using machine learning. It paves way for further investigation and realize a comprehensive security framework to protect CPS from internal and external cyber-attacks.

The first thing addressed in this paper is Plant fruit recognition and its feature extraction, it plays vital role in Agriculture. The point is to construct an exact, quick and solid framework utilizing CNN realities. Automatic fruit detection may reduce human efforts (counting number of fruits and manual identification). The second thing addressed in this paper is plant classification to avoid similarities in many plant fruits. Fruit classification may help fruit sellers to identify and to differentiate various kind of fruits having same similarities. The proposed framework has applied convolutional Neural Net (CNN) to the undertakings of distinguishing natural fruit pictures. In any case, deep learning has been shown as of late to be an extremely incredible picture identification procedure, and CNN is a best in class way to deal with deep learning.

Cotton detection is the localization and identification of the cotton in an image. It has a wide application in robot harvesting. Various modern algorithms use deep learning techniques for detection of fruits/flowers. As per the survey, the topics travelled include numerous algorithms used, and accuracy obtained on using those algorithms on their data set. The limitations and the advantages in each paper, are also discussed. This paper focuses on various fruit detection algorithms- the Faster RCNN, the RCNN, YOLO. Ultimately, a rigorous survey of many papers related to the detection of objects like fruits/flowers, analysis of the assets and faintness of each paper leads us to understanding the techniques and purpose of algorithms.

The excellent way to enhance the lives of the people with disabilities is technology. There have been several avenues of research that has opened the room after the advent of Artificial Intelligence. The Lives of people with disabilities can be improved by Artificial Intelligence. The smart devices have been supporting the people with sight loss and physical disabilities. Route planning software, ranging from robotic arms and other prosthetic limbs has made true advances which is a significant contribution done by Artificial Intelligence in the lives of persons with disability. The dependency of human life on technology is outreaching even to the Person with Disabilities. The prevailing technological trends have started becoming a current favorite amongst those who need assistance because of any form of Disability. While some artificially intelligent technology is still fairly new the potential is all there. The Paper discusses how Artificial Intelligence has helped in improving the living standards of Person with Disability, its future possibilities, potential legal issues concerning the implication of Artificial Intelligence in India. In this paper the authors want to put an emphasis on the ways by which life of persons with disability can be easier and swift in respect of their day-to-day activities by using modern technology namely Artificial Intelligence.

Available data in machine learning applications is becoming increasingly complex, due to higher dimensionality and difficult classes. There exists a wide variety of approaches to measuring complexity of labeled data, according to class overlap, separability or boundary shapes, as well as group morphology. Many techniques can transform the data in order to find better features, but few focus on specifically reducing data complexity. Most data transformation methods mainly treat the dimensionality aspect, leaving aside the available information within class labels which can be useful when classes are somehow complex. This paper proposes an autoencoder-based approach to complexity reduction, using class labels in order to inform the loss function about the adequacy of the generated variables. This leads to three different new feature learners, Scorer, Skaler and Slicer. They are based on Fisher's discriminant ratio, the Kullback-Leibler divergence and least-squares support vector machines, respectively. They can be applied as a preprocessing stage for a binary classification problem. A thorough experimentation across a collection of 27 datasets and a range of complexity and classification metrics shows that class-informed autoencoders perform better than 4 other popular unsupervised feature extraction techniques, especially when the final objective is using the data for a classification task., This paper has been accepted for publication by IEEE Transactions on Pattern Analysis and Machine Intelligence

Deep learning models have been shown to be vulnerable to adversarial attacks. Adversarial attacks are imperceptible perturbations added to an image such that the deep learning model misclassifies the image with a high confidence. Existing adversarial defenses validate their performance using only the classification accuracy. However, classification accuracy by itself is not a reliable metric to determine if the resulting image is "adversarial-free". This is a foundational problem for online image recognition applications where the ground-truth of the incoming image is not known and hence we cannot compute the accuracy of the classifier or validate if the image is "adversarial-free" or not. This paper proposes a novel privacy preserving framework for defending Black box classifiers from adversarial attacks using an ensemble of iterative adversarial image purifiers whose performance is continuously validated in a loop using Bayesian uncertainties. The proposed approach can convert a single-step black box adversarial defense into an iterative defense and proposes three novel privacy preserving Knowledge Distillation (KD) approaches that use prior meta-information from various datasets to mimic the performance of the Black box classifier. Additionally, this paper proves the existence of an optimal distribution for the purified images that can reach a theoretical lower bound, beyond which the image can no longer be purified. Experimental results on six public benchmark datasets namely: 1) Fashion-MNIST, 2) CIFAR-10, 3) GTSRB, 4) MIO-TCD, 5) Tiny-ImageNet, and 6) MS-Celeb show that the proposed approach can consistently detect adversarial examples and purify or reject them against a variety of adversarial attacks.

The introduction of deep learning technology can improve the performance of cyber-physical systems (CPSs) in many ways. However, this also brings new security issues. To tackle these challenges, this paper explores the vulnerabilities of deep learning-based unmanned aerial vehicles (UAVs), which are typical CPSs. Although many research works have been reported previously on adversarial attacks of deep learning models, only few of them are concerned about safety-critical CPSs, especially regression models in such systems. In this paper, we analyze the problem of adversarial attacks against deep learning-based UAVs and propose two adversarial attack methods against regression models in UAVs. Experiments demonstrate that the proposed non-targeted and targeted attack methods both can craft imperceptible adversarial images and pose a considerable threat to the navigation and control of UAVs. To address this problem, adversarial training and defensive distillation methods are further investigated and evaluated, increasing the robustness of deep learning models in UAVs. To our knowledge, this is the first study on adversarial attacks and defenses against deep learning-based UAVs, which calls for more attention to the security and safety of such safety-critical applications.

Linear Regression (LR) is a classical machine learning algorithm which has many applications in the cyber physical social systems (CPSS) to shape and simplify the way we live, work and communicate. This paper focuses on the data analysis for CPSS when the Linear Regression is applied. The training process of LR is time-consuming since it involves complex matrix operations, especially when it gets a large scale training dataset In the CPSS. Thus, how to enable devices to efficiently perform the training process of the Linear Regression is of significant importance. To address this issue, in this paper, we present a secure, verifiable and fair approach to outsource LR to an untrustworthy cloud-server. In the proposed scheme, computation inputs/outputs are obscured so that the privacy of sensitive information is protected against cloud-server. Meanwhile, computation result from cloud-server is verifiable. Also, fairness is guaranteed by the blockchain, which ensures that the cloud gets paid only if he correctly performed the outsourced workload. Based on the presented approach, we exploited the fair, secure outsourcing system on the Ethereum blockchain. We analysed our presented scheme on theoretical and experimental, all of which indicate that the presented scheme is valid, secure and efficient.

Reinforcement learning (RL) has demonstrated its ability to solve high dimensional tasks by leveraging non-linear function approximators. However, these successes are mostly achieved by 'black-box' policies in simulated domains. When deploying RL to the real world, several concerns regarding the use of a 'black-box' policy might be raised. In order to make the learned policies more transparent, we propose in this paper a policy iteration scheme that retains a complex function approximator for its internal value predictions but constrains the policy to have a concise, hierarchical, and human-readable structure, based on a mixture of interpretable experts. Each expert selects a primitive action according to a distance to a prototypical state. A key design decision to keep such experts interpretable is to select the prototypical states from trajectory data. The main technical contribution of the paper is to address the challenges introduced by this non-differentiable prototypical state selection procedure. Experimentally, we show that our proposed algorithm can learn compelling policies on continuous action deep RL benchmarks, matching the performance of neural network based policies, but returning policies that are more amenable to human inspection than neural network or linear-in-feature policies.

Mutation analysis assesses a test suites adequacy by measuring its ability to detect small artificial faults, systematically seeded into the tested program. Mutation analysis is considered one of the strongest test-adequacy criteria. Mutation testing builds on top of mutation analysis and is a testing technique that uses mutants as test goals to create or improve a test suite. Mutation testing has long been considered intractable because the sheer number of mutants that can be created represents an insurmountable problemboth in terms of human and computational effort. This has hindered the adoption of mutation testing as an industry standard. For example, Google has a codebase of two billion lines of code and more than 150,000,000 tests are executed on a daily basis. The traditional approach to mutation testing does not scale to such an environment; even existing solutions to speed up mutation analysis are insufficient to make it computationally feasible at such a scale. To address these challenges, this paper presents a scalable approach to mutation testing based on the following main ideas: (1) mutation testing is done incrementally, mutating only changed code during code review, rather than the entire code base; (2) mutants are filtered, removing mutants that are likely to be irrelevant to developers, and limiting the number of mutants per line and per code review process; (3) mutants are selected based on the historical performance of mutation operators, further eliminating irrelevant mutants and improving mutant quality. This paper empirically validates the proposed approach by analyzing its effectiveness in a code-review-based setting, used by more than 24,000 developers on more than 1,000 projects. The results show that the proposed approach produces orders of magnitude fewer mutants and that context-based mutant filtering and selection improve mutant quality and actionability. Overall, the proposed approach represents a mutation testing framework that seamlessly integrates into the software development workflow and is applicable to industrial settings of any size.

Diabetic Retinopathy (DR) is the condition caused due to uncontrolled diabetes that can lead to vision impairment. It greatly affects the retinal blood vessels and diminishes the fundus light-sensitive inner coating. Early diagnosis and regular screening of this disease are essential for prompt processing through artificial intelligence techniques. This paper targets assessing the latest techniques for screening and diagnosing DR, including 94 articles based on the Detection and grading of DR. For every analyzed approach, tables are summarized detailing imaging procedure used, datasets, performance metrics used. The research gaps are also highlighted in this paper. Despite the consistent progression and methods actualized in this field, a couple of issues actually should be centered on. The noise and contrast of the image in Image enhancement are still in the infancy stage for high resolution. This study covers a review of existing image techniques, the gold standard and private datasets available, performance measures used for detection and grading of DR. Now the future research focuses on the amalgamation of the dataset as well as techniques to make the generalized technique for detecting the lesion in DR through an automated system. Moreover, various research gaps have also been taken into account for further research. This review is beneficial to the researchers working in the field of medical imaging to screen and diagnose diseases.

This paper serves as a survey of recent advances in large margin training and its theoretical foundations, mostly for (nonlinear) deep neural networks (DNNs) that are probably the most prominent machine learning models for large-scale data in the community over the past decade. We generalize the formulation of classification margins from classical research to latest DNNs, summarize theoretical connections between the margin, network generalization, and robustness, and introduce recent efforts in enlarging the margins for DNNs comprehensively. Since the viewpoint of different methods is discrepant, we categorize them into groups for ease of comparison and discussion in the paper. Hopefully, our discussions and overview inspire new research work in the community that aim to improve the performance of DNNs, and we also point to directions where the large margin principle can be verified to provide theoretical evidence why certain regularizations for DNNs function well in practice. We managed to shorten the paper such that the crucial spirit of large margin learning and related methods are better emphasized., Accepted by TPAMI, 8 pages, 3 figures

Security issues in mobile apps are increasingly relevant as this software have become part of the daily life of billions of people. As the dominant OS, Android is a primary target for ill-intentioned programmers willing to exploit its vulnerabilities by spreading malwares. Significant research has been devoted to the identification of these malwares. The current paper is an extension of our previous effort to contribute to said research with a new benchmark of Android vulnerabilities. We proposed AndroVul, a repository for Android security vulnerabilities, that builds on AndroZoo – a well-known Android app dataset – and contains data on vulnerabilities for a representative sample of about 16,000 Android apps. The present paper adds confirmed malwares from the VirusShare dataset and explores more thoroughly the effectiveness of different machine learning techniques, with respect to the classification of malicious apps. We investigated different classifiers and feature selection techniques as well as different combinations for our input data. Our results suggest that the classifier MPL is the leading classifier, with competitive results that favorably compare to recent malware detection work. Additionally, we investigate how to classify (as benign or malicious) AndroZoo apps based on the number of antivirus flags they are tagged with. We found that different thresholds only marginally affect the machine learning classifier results and that the strictest choice (i.e. one flag) performs best on the confirmed malwares from VirusShare.

Today, the rapid growth of the internet and advancements in mobile technology and increased internet connectivity have brought us to a data-driven economy where an enormous amount of data is being used to train machine learning models to make strategic decisions. However, in the aftermath of a data breach by Facebook in 2018, there are some serious concerns over user data privacy and security being used to train the Machine Learning models. In this context, a new approach, Federated Machine Learning is now one of the most talked-about and recent approaches. Current research primarily focuses on Federated Learning's advantages over the traditional methods and/or its classification. However, being in a nascent stage of development as a method, certain challenges need to be addressed. This paper intends to address the totality of federated learning with a complete vulnerability assessment. During the study of the literature, it is found that security being promised as one of the key advantages of federated learning can still not be guaranteed because of some issues inherently present, and this can lead to poisoning, inference attacks and insertion of backdoors, etc. This paper intends to provide a complete picture by giving an in-depth and comprehensive analysis of Federated Learning and its taxonomy. It also provides a detailed vulnerability assessment and highlights the challenges faced in the current setting and future research directions to make federated learning a more functional, robust and secure method to train machine learning models.

Automated detection of software vulnerabilities is a fundamental problem in software security. Existing program analysis techniques either suffer from high false positives or false negatives. Recent progress in Deep Learning (DL) has resulted in a surge of interest in applying DL for automated vulnerability detection. Several recent studies have demonstrated promising results achieving an accuracy of up to 95% at detecting vulnerabilities. In this paper, we ask, "how well do the state-of-the-art DL-based techniques perform in a real-world vulnerability prediction scenario?". To our surprise, we find that their performance drops by more than 50%. A systematic investigation of what causes such precipitous performance drop reveals that existing DL-based vulnerability prediction approaches suffer from challenges with the training data (e.g., data duplication, unrealistic distribution of vulnerable classes, etc.) and with the model choices (e.g., simple token-based models). As a result, these approaches often do not learn features related to the actual cause of the vulnerabilities. Instead, they learn unrelated artifacts from the dataset (e.g., specific variable/function names, etc.). Leveraging these empirical findings, we demonstrate how a more principled approach to data collection and model design, based on realistic settings of vulnerability prediction, can lead to better solutions. The resulting tools perform significantly better than the studied baseline: up to 33.57% boost in precision and 128.38% boost in recall compared to the best performing model in the literature. Overall, this paper elucidates existing DL-based vulnerability prediction systems' potential issues and draws a roadmap for future DL-based vulnerability prediction research. In that spirit, we make available all the artifacts supporting our results: https://git.io/Jf6IA., Comment: Under Review IEEE Transactions on Software Engineering

Sentiment analysis aims to classify a text according to sentimental polarities of people’s opinions, such as positive, negative, or neutral. While most of the studies focus on eliciting features from English text, the research on Arabic is limited due to the morphological and grammatical complexity of Arabic language. In this paper, we proposed an optimized sentiment classification for dialectal Arabic short text at the document level using deep learning (DL). The contributions of this paper are in three areas. First, we extracted semantic features for Arabic short text at the word level and character level. Second, we used three DL topologies for classification models: a long short-term memory recurrent neural network (LSTM); a convolutional neural network (CNN); and an ensemble model combining both models’ advantages to improve the prediction performance. Third, we used a hyperparameter tuning estimation method to optimize the neural network performance. We trained and tested our proposed models on a dataset that consists of Modern Standard Arabic and dialectal Arabic corpus collected from Twitter. The results showed significant improvement in Arabic text classification in term of classification accuracy that ranges between 88% and 69.7%. The ensemble model scored the highest accuracy of 96.7% on the test set.

Word Sense Disambiguation (WSD) is significant for improving the accuracy of the interpretation of a Natural language text. Various supervised learning-based models and knowledge-based models have been developed in the literature for WSD of the language text. However, these models do not provide good results for low-resource languages, due to the lack of labelled and tagged data. Therefore, in this work, we have examined different word embedding techniques for word sense disambiguation of the Hindi language texts. Several studies in the literature show that these embeddings have been utilized for different foreign languages in the field of word sense disambiguation. However, to the best of our knowledge, no such work exists for the Hindi language. Therefore, in this paper, we utilize various existing word embeddings for WSD of Hindi text. Moreover, we have created Hindi word embeddings on articles taken from Wikipedia and test the quality of the created word embeddings using Pearson correlation. In this direction, we perform different experiments and observe that Word2Vec model gives best performance among all the considered embeddings on the used Hindi dataset. In our method, the proposed model directly takes input that is trained with word embedding methods and helps to develop a sense inventory using clustering that has been employed for performing disambiguation. Experimental observations indicate that the performance of the proposed approach is moderate and competent in terms of accuracy. The paper, thus, presents how WSD can leverage these representations to encode rich semantic information.

In the past ten years, many high-quality datasets have been released to support the rapid development of deep learning in the fields of computer vision, voice, and natural language processing. Nowadays, deep learning has become a key research component of the Sixth-Generation wireless systems (6G) with numerous regulatory and defense applications. In order to facilitate the application of deep learning in radio signal recognition, in this work, a large-scale real-world radio signal dataset is created based on a special aeronautical monitoring system - Automatic Dependent Surveillance-Broadcast (ADS-B). This paper makes two main contributions. First, an automatic data collection and labeling system is designed to capture over-the-air ADS-B signals in the open and real-world scenario without human participation. Through data cleaning and sorting, a high-quality dataset of ADS-B signals is created for radio signal recognition. Second, we conduct an in-depth study on the performance of deep learning models using the new dataset, as well as comparison with a recognition benchmark using machine learning and deep learning methods. Finally, we conclude this paper with a discussion of open problems in this area.

In spite of the insight commonality may have concerning agrarian practice, fact is that nowadays agricultural science diligence is accurate, precise, data-driven, and vigorous than ever. The emanation of the technologies based on Internet of Things (IoT) has reformed nearly each industry like smart city, smart health, smart grid, smart home, including “smart agriculture or precision agriculture”. Applying machine learning using the IoT data analytics in agricultural sector will rise new benefits to increase the quantity and quality of production from the crop fields to meet the increasing food demand. Such world-shattering advancements are rocking the current agrarian approaches and generating novel and best chances besides a number of limitations. This paper climaxes the power and capability of computing techniques including internet of things, wireless sensor networks, data analytics and machine learning in agriculture. The paper proposed the prediction model of Apple disease in the apple orchards of Kashmir valley using data analytics and Machine learning in IoT system. Furthermore, a local survey was conducted to know from the farmers about the trending technologies and their effect in precision agriculture. Finally, the paper discusses the challenges faced when incorporating these technologies in the traditional farming approaches.

Localization forms the heart of various autonomous mobile robots. For efficient navigation, these robots need to adopt effective localization strategy. This paper, presents a comprehensive review on localization system, problems, principle and approaches for mobile robots. First, we classify the localization problems in to three categories based on the information of initial position of the robot. Next, we discuss on robot position update principles. Then, we discuss key techniques to localize the mobile robot such as: probabilistic approach, autonomous map building and radio frequency identification (RFID) based scheme. In the probabilistic localization section, we discuss the Markov localization and Kalman filter along with its extended versions. Autonomous map building focuses on the widely used simultaneous localization and mapping (SLAM) approach. This section also discusses on applying SLAM to localize brain-controlled mobile robots. Next, we discuss on applying evolutionary approaches to estimate optimal position. The RFID scheme addresses on effective utilization of RFID tags to track objects and position the robot. We then analyze on position and orientation errors occurred by different localization strategies. We conclude this paper by highlighting future research possibilities.