Computer-based Algorithms Research Articles

Automatic screening of cervical cells using block image processing.

BackgroundCervical cancer is the second leading cause of female-specific cancer-related deaths after breast cancer, especially in developing countries. However, the incidence of the disease may be significantly decreased if the patient is diagnosed in the pre-cancerous lesion stage or earlier. In recent years, computer-based algorithms are widely used in cervical cancer screening. Most of the proposed algorithms follow the procedure of segmentation, feature extraction, and then classification. Nevertheless, few of the existing segmentation methods are as flexible and robust as the human visual system, and the complexity of the algorithms makes it difficult for clinical application.MethodsIn this study, a computer-assisted analytical approach is proposed to identify the existence of suspicious cells in a whole slide cervical cell image (WSCCI). The main difference between our method and the conventional algorithm is that the image is divided into blocks with certain size instead of segmented cells, which can greatly reduce the computational complexity. Via data analysis, some texture and color histogram features show significant differences between blocks with and without suspicious cells. Therefore these features can be used as the input of the support vector machine classifier. 1100 non-background blocks (110 suspicious blocks) are trained to build a model, while 1040 blocks (491 non-background blocks) from 12 other WSCCIs are tested to verify the feasibility of the algorithm.ResultsThe experimental results show that the accuracy of our method is about 98.98 %. More importantly, the sensitivity, which is more fatal in cancer screening, is 95.0 % according to the images tested in the study, while the specificity is 99.33 %.ConclusionThe analysis of the algorithm is based on block images, which is different from conventional methods. Although some analysis work should be done in advance, the later processing speed will be greatly enhanced with the establishment of the model. Furthermore, since the algorithm is based on the actual WSCCI, the method will be of directive significance for clinical screening.

Automatic extraction and identification of users’ responses in Facebook medical quizzes

BackgroundIn the last few years the use of social media in medicine has grown exponentially, providing a new area of research based on the analysis and use of Web 2.0 capabilities. In addition, the use of social media in medical education is a subject of particular interest which has been addressed in several studies. One example of this application is the medical quizzes of The New England Journal of Medicine (NEJM) that regularly publishes a set of questions through their Facebook timeline. ObjectiveWe present an approach for the automatic extraction of medical quizzes and their associated answers on a Facebook platform by means of a set of computer-based methods and algorithms. MethodsWe have developed a tool for the extraction and analysis of medical quizzes stored on Facebook timeline at the NEJM Facebook page, based on a set of computer-based methods and algorithms using Java. The system is divided into two main modules: Crawler and Data retrieval. ResultsThe system was launched on December 31, 2014 and crawled through a total of 3004 valid posts and 200,081 valid comments. The first post was dated on July 23, 2009 and the last one on December 30, 2014. 285 quizzes were analyzed with 32,780 different users providing answers to the aforementioned quizzes. Of the 285 quizzes, patterns were found in 261 (91.58%). From these 261 quizzes where trends were found, we saw that users follow trends of incorrect answers in 13 quizzes and trends of correct answers in 248. ConclusionsThis tool is capable of automatically identifying the correct and wrong answers to a quiz provided on Facebook posts in a text format to a quiz, with a small rate of false negative cases and this approach could be applicable to the extraction and analysis of other sources after including some adaptations of the information on the Internet.

The use of facial modeling and analysis to objectively quantify facial redness.

The reproducible evaluation of facial redness is critical to the assessment of erythematotelangiectatic rosacea. Assessments have typically focused on the use of photography with the use of semi-quantitative grading scales based on evaluator rating. However, few studies have utilized computer-based algorithms to evaluate facial redness. The purpose of this clinical study was to assess whether there is correlation between clinical grading of facial redness to the assessment of a quantitative computer-based facial modeling and measurement. In this prospective study, a set of high-resolution facial photographs and cross-polarized subsurface photographs for erythema detection were obtained for 31 study participants. A computer algorithm was then utilized to detect and quantify facial redness in the photographs and compare this to semi-quantitative evaluator-based grading for facial redness. There was a strong correlation between computer-based cross-polarized subsurface erythema quantification and clinical grading for redness intensity (Clinical Erythema Assessment), redness distribution, and overall redness severity (Modified Clinical Erythema Assessment). Overall, facial redness measurements by facial imaging and computer analysis correlated well to clinical grading scales for both redness intensity and distribution. Future studies should incorporate facial modeling and analysis tools for assessments in clinical studies to introduce greater objectivity and quantitative analysis in facial erythema-based analyses.

Abstracts from ATTD 20158th International Conference on Advanced Technologies & Treatments for DiabetesParis, France—February 18–21, 2015

Abstracts from ATTD 20158th International Conference on Advanced Technologies & Treatments for DiabetesParis, France—February 18–21, 2015

Validating estimates of prevalence of non-communicable diseases based on household surveys: the symptomatic diagnosis study.

BackgroundEasy-to-collect epidemiological information is critical for the more accurate estimation of the prevalence and burden of different non-communicable diseases around the world. Current measurement is restricted by limitations in existing measurement systems in the developing world and the lack of biometry tests for non-communicable diseases. Diagnosis based on self-reported signs and symptoms (“Symptomatic Diagnosis,” or SD) analyzed with computer-based algorithms may be a promising method for collecting timely and reliable information on non-communicable disease prevalence. The objective of this study was to develop and assess the performance of a symptom-based questionnaire to estimate prevalence of non-communicable diseases in low-resource areas.MethodsAs part of the Population Health Metrics Research Consortium study, we collected 1,379 questionnaires in Mexico from individuals who suffered from a non-communicable disease that had been diagnosed with gold standard diagnostic criteria or individuals who did not suffer from any of the 10 target conditions. To make the diagnosis of non-communicable diseases, we selected the Tariff method, a technique developed for verbal autopsy cause of death calculation. We assessed the performance of this instrument and analytical techniques at the individual and population levels.ResultsThe questionnaire revealed that the information on health care experience retrieved achieved 66.1% (95% uncertainty interval [UI], 65.6–66.5%) chance corrected concordance with true diagnosis of non-communicable diseases using health care experience and 0.826 (95% UI, 0.818–0.834) accuracy in its ability to calculate fractions of different causes. SD is also capable of outperforming the current estimation techniques for conditions estimated by questionnaire-based methods.ConclusionsSD is a viable method for producing estimates of the prevalence of non-communicable diseases in areas with low health information infrastructure. This technology can provide higher-resolution prevalence data, more flexible data collection, and potentially individual diagnoses for certain conditions.Electronic supplementary materialThe online version of this article (doi:10.1186/s12916-014-0245-8) contains supplementary material, which is available to authorized users.

Generation of cytotoxic T lymphocytes specific for native or modified peptides derived from the epidermal growth factor receptor pathway substrate 8 antigen.

The ideal tumor antigen for the development of a cancer immunotherapy is one that is expressed only in tumor cells. The epidermal growth factor receptor pathway substrate 8 gene (Eps8) might be an effective antigen for cancer immunotherapy as it is overexpressed in a variety of cancer cells but not in normal tissues. In this study, the potential utility of an Eps8-derived immunotherapy was tested in vitro and in vivo. Three computer-based algorithms were used to design eight Eps8 native epitopes with potentially high binding affinity to the HLA-A2.1 molecule, which is found at a high frequency in the Chinese population. Of these eight, three peptides with a moderate affinity to the HLA-A2.1 molecule were modified at anchor residue positions to achieve stronger immunogenicity. These four modified peptides displayed stronger binding affinity to HLA-A2.1 molecules on T2 cells and a lower dissociation rate. In functional assays with human PBMCs in vitro and in HLA-A2.1/K(b) transgenic mice in vivo, CTLs primed by each native and modified peptide secreted IFN-γ and were toxic to cancer cells from a variety of tissue types in an HLA-A2.1-restricted and Eps8-specific manner. p101-109-2L and p276-284-1Y9V were superior to other modified and native epitopes both in vitro and in vivo. These results indicate that employing the native and modified epitopes identified here in Eps8-based immunotherapy for HLA-A2.1 positive cancer patients may result in efficient anticancer immune responses for diverse tumor types.

Relational machine learning for electronic health record-driven phenotyping

ObjectiveElectronic health records (EHR) offer medical and pharmacogenomics research unprecedented opportunities to identify and classify patients at risk. EHRs are collections of highly inter-dependent records that include biological, anatomical, physiological, and behavioral observations. They comprise a patient’s clinical phenome, where each patient has thousands of date-stamped records distributed across many relational tables. Development of EHR computer-based phenotyping algorithms require time and medical insight from clinical experts, who most often can only review a small patient subset representative of the total EHR records, to identify phenotype features. In this research we evaluate whether relational machine learning (ML) using inductive logic programming (ILP) can contribute to addressing these issues as a viable approach for EHR-based phenotyping. MethodsTwo relational learning ILP approaches and three well-known WEKA (Waikato Environment for Knowledge Analysis) implementations of non-relational approaches (PART, J48, and JRIP) were used to develop models for nine phenotypes. International Classification of Diseases, Ninth Revision (ICD-9) coded EHR data were used to select training cohorts for the development of each phenotypic model. Accuracy, precision, recall, F-Measure, and Area Under the Receiver Operating Characteristic (AUROC) curve statistics were measured for each phenotypic model based on independent manually verified test cohorts. A two-sided binomial distribution test (sign test) compared the five ML approaches across phenotypes for statistical significance. ResultsWe developed an approach to automatically label training examples using ICD-9 diagnosis codes for the ML approaches being evaluated. Nine phenotypic models for each ML approach were evaluated, resulting in better overall model performance in AUROC using ILP when compared to PART (p=0.039), J48 (p=0.003) and JRIP (p=0.003). DiscussionILP has the potential to improve phenotyping by independently delivering clinically expert interpretable rules for phenotype definitions, or intuitive phenotypes to assist experts. ConclusionRelational learning using ILP offers a viable approach to EHR-driven phenotyping.

Glucose control in the ICU: is there a time for more ambitious targets again?

During the last 2 decades, the treatment of hyperglycemia in critically ill patients has become one of the most discussed topics in the intensive medicine field. The initial data suggesting significant benefit of normalization of blood glucose levels in critically ill patients using intensive intravenous insulin therapy have been challenged or even neglected by some later studies. At the moment, the need for glucose control in critically ill patients is generally accepted yet the target glucose values are still the subject of ongoing debates. In this review, we summarize the current data on the benefits and risks of tight glucose control in critically ill patients focusing on the novel technological approaches including continuous glucose monitoring and its combination with computer-based algorithms that might help to overcome some of the hurdles of tight glucose control. Since increased risk of hypoglycemia appears to be the major obstacle of tight glucose control, we try to put forward novel approaches that may help to achieve optimal glucose control with low risk of hypoglycemia. If such approaches can be implemented in real-world practice the entire concept of tight glucose control may need to be revisited.

Genetics of differential susceptibility to pneumonic Burkholderia pseudomallei infection (INC7P.424)

Abstract Burkholderia pseudomallei (Bp) is a pathogenic gram-negative bacterium that causes the severe human disease melioidosis. Bp is listed as a CDC Tier 1 select agent and is considered a potential bioterrorism agent that poses a threat to national security if intentionally released into the human population. Our goal is to use a murine model of Bp infection to gain a better understanding of the interactions between this bacterium and its hosts. We have utilized fully genotyped recombinant inbred BXD mice and a powerful array of complementary computer-based modeling algorithms and databases collectively known as the GeneNetwork. Preliminary studies revealed that Bp infection elicits phenotypically distinct innate immune responses in terms of immune cell recruitment to the lungs, survival and weight loss following pneumonic infection in parental and BXD mice. Preliminary interval mapping of our survival and weight loss phenotypic data using GeneNetwork revealed that survival is a complex trait involving loci on chromosomes 5, 7, and 9 and weight retention involves loci on chromosome 12. Furthermore, we have identified several potential candidate genes within the significant and/or suggestive interval on these chromosomes that appear to correlate with differential susceptibility to Bp infection. These results form the foundation for future work that will significantly increase our understanding of the interactions between Bp and its genetically diverse hosts.

Modelling induced resistance to plant diseases

Plant disease control has traditionally relied heavily on the use of agrochemicals despite their potentially negative impact on the environment. An alternative strategy is that of induced resistance (IR). However, while IR has proven effective in controlled environments, it has shown variable field efficacy, thus raising questions about its potential for disease management in a given crop. Mathematical modelling of IR assists researchers with understanding the dynamics of the phenomenon in a given plant cohort against a selected disease-causing pathogen. Here, a prototype mathematical model of IR promoted by a chemical elicitor is proposed and analysed. Standard epidemiological models describe that, under appropriate environmental conditions, Susceptible plants (S) may become Diseased (D) upon exposure to a compatible pathogen or are able to Resist the infection (R) via basal host defence mechanisms. The application of an elicitor enhances the basal defence response thereby affecting the relative proportion of plants in each of the S, R and D compartments. IR is a transient response and is modelled using reversible processes to describe the temporal evolution of the compartments. Over time, plants can move between these compartments. For example, a plant in the R-compartment can move into the S-compartment and can then become diseased. Once in the D-compartment, however, it is assumed that there is no recovery. The terms in the equations are identified using established principles governing disease transmission and this introduces parameters which are determined by matching data to the model using computer-based algorithms. These then give the best match of the model with experimental data. The model predicts the relative proportion of plants in each compartment and quantitatively estimates elicitor effectiveness. An illustrative case study will be given; however, the model is generic and will be applicable for a range of plant–pathogen–elicitor scenarios.

Structural Modeling and Physicochemical Characterization Provide Evidence that P66 Forms a β-Barrel in the Borrelia burgdorferi Outer Membrane

The Borrelia burgdorferi outer membrane (OM) contains numerous surface-exposed lipoproteins but a relatively low density of integral OM proteins (OMPs). Few membrane-spanning OMPs of B. burgdorferi have been definitively identified, and none are well characterized structurally. Here, we provide evidence that the borrelial OMP P66, a known adhesin with pore-forming activity, forms a β-barrel in the B. burgdorferi OM. Multiple computer-based algorithms predict that P66 forms a β-barrel with either 22 or 24 transmembrane domains. According to our predicted P66 topology, a lysine residue (K487) known to be sensitive to trypsin cleavage is located within a surface-exposed loop. When we aligned the mature P66 amino acid sequences from B. burgdorferi and B. garinii, we found that K487 was present only in the B. burgdorferi P66 protein sequence. When intact cells from each strain were treated with trypsin, only B. burgdorferi P66 was trypsin sensitive, indicating that K487 is surface exposed, as predicted. Consistent with this observation, when we inserted a c-Myc tag adjacent to K487 and utilized surface localization immunofluorescence, we detected the loop containing K487 on the surface of B. burgdorferi. P66 was examined by both Triton X-114 phase partitioning and circular dichroism, confirming that the protein is amphiphilic and contains extensive (48%) β-sheets, respectively. Moreover, P66 also was able to incorporate into liposomes and form channels in large unilamellar vesicles. Finally, blue native PAGE (BN-PAGE) revealed that under nondenaturing conditions, P66 is found in large complexes of ∼400 kDa and ∼600 kDa. Outer surface lipoprotein A (OspA) and OspB both coimmunoprecipitate with P66, demonstrating that P66 associates with OspA and OspB in B. burgdorferi. The combined computer-based structural analyses and supporting physicochemical properties of P66 provide a working model to further examine the porin and integrin-binding activities of this OMP as they relate to B. burgdorferi physiology and Lyme disease pathogenesis.

Using meta-regression data mining to improve predictions of performance based on heart rate dynamics for Australian football

This work investigates the effectiveness of using computer-based machine learning regression algorithms and meta-regression methods to predict performance data for Australian football players based on parameters collected during daily physiological tests. Three experiments are described. The first uses all available data with a variety of regression techniques. The second uses a subset of features selected from the available data using the Random Forest method. The third used meta-regression with the selected feature subset. Our experiments demonstrate that feature selection and meta-regression methods improve the accuracy of predictions for match performance of Australian football players based on daily data of medical tests, compared to regression methods alone. Meta-regression methods and feature selection were able to obtain performance prediction outcomes with significant correlation coefficients. The best results were obtained by the additive regression based on isotonic regression for a set of most influential features selected by Random Forest. This model was able to predict athlete performance data with a correlation coefficient of 0.86 (p<0.05).

Nonverbal Communication in Medieval Illustrations Revisited by Computer Vision and Art History

This contribution discusses how computer vision combined with art history can analyze the visual codes and artistic representations of embodied communication represented in medieval culture. Our computer-based detection algorithms are able to search directly for gestures in images and thus avoid limitations of retrieval systems that search only textual annotations. As a result, art history is provided not only with a system that enables efficient access to large image datasets, but also a quantitative analysis of the variability and interrelation among gestures or between gestures and other objects. We base our approach on one of four illustrated manuscripts of Eike von Repgow's (ca. 1180–ca. 1235) Sachsenspiegel (Mirror of the Saxons), which reveals a visual grammar that arranges the gestures to a certain extent, but within that framework, the drafter composes freely and with artistic perspective.

SIMULATION OF CLOSE-RANGE PHOTOGRAMMETRIC SYSTEMS FOR INDUSTRIAL SURFACE INSPECTION

Abstract. Close-range photogrammetric measurement systems are increasingly used for high-precision surface inspection of car body parts. These measurement systems are based on an active light source, the projector, and one or more cameras. Many systems use a sequence of fringe projection, mostly a combination of the gray code and phase shift technique. Basically the quality of the measurement result depends on best possible positions of these sensors, which requires human expert knowledge and experience. But is it possible to use computer-based algorithms to find optimal measuring positions? Simulation processes are discovered as part of a research project aimed at the evaluation of the quality of measuring positions concerning to visibility, the attainable accuracy and realizable feature extraction. One approach is the simulation of the photogrammetric sensor using ray tracing techniques to create photorealistic pictures from the sensor cameras view. This image sequence could be processed with the evaluation software of the system manufacturer in order to calculate a three dimensional point cloud. Following an actual/target comparison should indicate differences that trace back to insufficient measuring positions. In this paper we show how to build up a virtual close range photogrammetric sensor using POV Ray, a free ray tracing software. After introducing the simulation concept, the design of a virtual close range photogrammetric sensor is presented. Based on practical examples of sampled scenes the potential of photorealistic ray tracing is shown. Finally the usability of this simulation approach is discussed.

Modeling induced resistance to plant disease using a dynamical systems approach

Modeling induced resistance to plant disease using a dynamical systems approach

Antiviral Stratagems against HIV-1 Using RNA Interference (RNAi) Technology

The versatility of human immunodeficiency virus (HIV)-1 and its evolutionary potential to elude antiretroviral agents by mutating may be its most invincible weapon. Viruses, including HIV, in order to adapt and survive in their environment evolve at extremely fast rates. Given that conventional approaches which have been applied against HIV have failed, novel and more promising approaches must be employed. Recent studies advocate RNA interference (RNAi) as a promising therapeutic tool against HIV. In this regard, targeting multiple HIV sites in the context of a combinatorial RNAi-based approach may efficiently stop viral propagation at an early stage. Moreover, large high-throughput RNAi screens are widely used in the fields of drug development and reverse genetics. Computer-based algorithms, bioinformatics, and biostatistical approaches have been employed in traditional medicinal chemistry discovery protocols for low molecular weight compounds. However, the diversity and complexity of RNAi screens cannot be efficiently addressed by these outdated approaches. Herein, a series of novel workflows for both wet- and dry-lab strategies are presented in an effort to provide an updated review of state-of-the-art RNAi technologies, which may enable adequate progress in the fight against the HIV-1 virus.

The Effect of a Provider-Enhanced Clinical Decision Support Tool for Guiding Venous Thromboembolism Pharmacoprophylaxis in Low-Risk Patients

Background: Computer-based tools to assess venous thromboembolism (VTE) risk have been shown to increase VTE pharmacoprophylaxis rates and decrease VTE incidence in high-risk hospitalized patients. However, VTE risk may be underestimated using computer-based tools alone. We tested the effect of a provider-enhanced clinical decision support (CDS) tool on VTE pharmacoprophylaxis and VTE incidence in patients who would have been deemed “low risk” using a computer-based risk-assessment tool alone. Methods: The study sample was adult patients hospitalized during a 13-month period who were determined to be at low risk for VTE with a computer-based risk-assessment tool. The provider-enhanced CDS tool was implemented 4 months into the study period and required providers to stratify patients as being at high, moderate, or low risk for VTE. We compared rates of VTE pharmacoprophylaxis and VTE incidence before and after implementation of the provider-enhanced CDS tool. Results: There were 1322 patients in the 4-month pre-implementation period and 3347 patients in the 9-month post-implementation period who were determined to be at low risk for VTE based on a computer-based risk-assessment tool. Using the provider-enhanced CDS tool, providers stratified 31% of these computer-assigned low-risk patients as being at moderate risk for VTE and 7% as being at high risk for VTE. The rate of VTE pharmacoprophylaxis increased from 27% to 34% (P < 0.01). The venous thromboprophylaxis rate decreased from 0.98% to 0.42% after implementation of the provider-enhanced CDS tool (P < 0.02). Conclusion: Patients identified as being at low risk for VTE solely by computer-based algorithms may include patients whom providers determine to be at risk for VTE. Provider-enhanced CDS tools may outperform computer-based VTE risk-stratification algorithms.

Programming from the bottom up [Books

In the first book review, &#x0201C;Programming from the Bottom Up,&#x0201D;, Norman Chonacky reviews an introductory textbook that considers fundamental computer operations as well as the methods and practices needed to create, use, or manage programs intended for the computational sciences and engineering. In the second book review, &#x0201C;Can Artificial Intelligence Improve Human Reasoning?&#x0201D;' George F. Luger reviews a book dealing with how the rational human mind might be understood and enhanced through computer-based representations and algorithms.

The Role of Item Models in Automatic Item Generation

Automatic item generation represents a relatively new but rapidly evolving research area where cognitive and psychometric theories are used to produce tests that include items generated using computer technology. Automatic item generation requires two steps. First, test development specialists create item models, which are comparable to templates or prototypes, that highlight the features or elements in the assessment task that must be manipulated. Second, these item model elements are manipulated to generate new items with the aid of computer-based algorithms. With this two-step process, hundreds or even thousands of new items can be created from a single item model. The purpose of our article is to describe seven different but related topics that are central to the development and use of item models for automatic item generation. We start by defining item model and highlighting some related concepts; we describe how item models are developed; we present an item model taxonomy; we illustrate how item models can be used for automatic item generation; we outline some benefits of using item models; we introduce the idea of an item model bank; and finally, we demonstrate how statistical procedures can be used to estimate the parameters of the generated items without the need for extensive field or pilot testing.

Analysis of carotid artery plaque and wall boundaries on CT images by using a semi-automatic method based on level set model

The purpose of this study was to evaluate the potentialities of a semi-automated technique in the detection and measurement of the carotid artery plaque. Twenty-two consecutive patients (18 males, 4 females; mean age 62years) examined with MDCTA from January 2011 to March 2011 were included in this retrospective study. Carotid arteries are examined with a 16-multi-detector-row CT system, and for each patient, the most diseased carotid was selected. In the first phase, the carotid plaque was identified and one experienced radiologist manually traced the inner and outer boundaries by using polyline and radial distance method (PDM and RDM, respectively). In the second phase, the carotid inner and outer boundaries were traced with an automated algorithm: level-set-method (LSM). Data were compared by using Pearson rho correlation, Bland-Altman, and regression. A total of 715 slices were analyzed. The mean thickness of the plaque using the reference PDM was 1.86mm whereas using the LSM-PDM was 1.96mm; using the reference RDM was 2.06mm whereas using the LSM-RDM was 2.03mm. The correlation values between the references, the LSM, the PDM and the RDM were 0.8428, 0.9921, 0.745 and 0.6425. Bland-Altman demonstrated a very good agreement in particular with the RDM method. Results of our study indicate that LSM method can automatically measure the thickness of the plaque and that the best results are obtained with the RDM. Our results suggest that advanced computer-based algorithms can identify and trace the plaque boundaries like an experienced human reader.