Search

Multi-task learning is becoming more popular and is being applied in a variety of applications. It improves the accuracy of prediction by simultaneously learning related tasks and saves cost through shared structures. In particular, the prediction of event type from social data is also an area where multi-task learning can be utilized. In this paper, we present a novel deep learning framework called S patial E vents P rediction (SEP) based on multi-task learning to predict the types of events that happen at a specific location from social data. The proposed model focuses on predicting the attribute types of an event, which is referred to as subtypes. Specifically, an event type-specific attention mechanism is introduced to extract the representations of social data and to identify their important components. The proposed attention mechanism is based on a two-level attention, which measures the importance of words and sentences to the subtypes of an event. We also propose a representation sharing method using semantic and spatial relationships between locations to alleviate the sparsity and incompleteness of data. The proposed representation sharing preserves the spatial heterogeneity between locations and significantly improves the accuracy of the overall framework. Experiments with real-world datasets confirm the effectiveness and efficiency of the proposed method.

Nowadays, question answering (QA) over complex questions is the most spotlighted research topic in diverse communities. Different from existing QA approaches for simple questions that require a single relation in a knowledge graph (KG) to find an answer set, QA for complex questions requires to process multiple KG relations. Complex questions often include some representations for constraints (ordinal, temporal, etc.) that restrict an answer set for the given question. Despite lots of efforts to process such questions, there are still limitations in processing constraints, multiple relations, and variables. To solve the issues, we propose a novel QA method that first decomposes an input question and then generates a correct query graph with fully complete semantics. In order to fill with lossy semantics caused by the decomposition task, the proposed method employs Bi-GRU based model. The model integrates individual compositional semantics of sub-query graphs matched to decomposed sub-questions. Experimental results show the best performance compared to the state-of-art on complex questions.

Generally, QA systems suffer from the structural difference where a question is composed of unstructured data, while its answer is made up of structured data in a Knowledge Graph (KG). To bridge this gap, most approaches use lexicons to cover data that are represented differently. However, the existing lexicons merely deal with representations for entity and relation mentions rather than consulting the comprehensive meaning of the question. To resolve this, we design a novel predicate constraints lexicon which restricts subject and object types for a predicate. It facilitates a comprehensive validation of a subject, predicate and object simultaneously. In this paper, we propose Predicate Constraints based Question Answering (PCQA). Our method prunes inappropriate entity/relation matchings to reduce search space, thus leading to an improvement of accuracy. Unlike the existing QA systems, we do not use any templates but generates query graphs to cover diverse types of questions. In query graph generation, we put more focus on matching relations rather than linking entities. This is well-suited to the use of predicate constraints. Our experimental results prove the validity of our approach and demonstrate a reasonable performance compared to other methods which target WebQuestions and Free917 benchmarks.

Context-aware recommender systems (CARS) exploit multiple contexts to improve user experience in embracing new information and services. Tensor factorization (TF), a type of latent factor model, has achieved remarkable performance in CARS. TF learns latent representations of contexts by decomposing an observed rating tensor and combines the latent representations as a vector form to represent contextual influence on users and items. However, due to the limitation of the contextual expression power, they have difficulties in effectively capturing complex correlations among multiple contexts, and also the meaning of each context is diluted. To address the issue, we propose a reliable TF-based recommender system based on a proposed context tensor (CT-CARS), which incorporates a variety of correlations among contexts. CT-CARS contains a novel recommendation rating function and a learning algorithm. Specifically, the proposed context tensor elaborately captures the influences of both individual contexts and context combinations. Moreover, we introduce a novel parameter initialization based on past-learned results to improve the reliability of recommendations. The overall time complexity of our parameter learning algorithm grows linearly as dataset size increases. Experiments on six real-world datasets including two large-scaled datasets show that CT-CARS outperforms the existing state-of-the-art models in terms of both accuracy and reliability.

We determined the cost-effectiveness of two different diagnostic imaging strategies in guiding curative treatment of early-stage hepatocellular carcinoma (HCC).We developed a decision analytic model using as its starting point a cohort of patients aged 55 years with early-stage HCC detected at dynamic multiphasic CT and with Child-Pugh class A cirrhosis. The model compared two strategies on the initial workup: conventional CT strategy using dynamic multiphasic CT only and gadoxetic acid-enhanced MRI strategy using additional gadoxetic acid-enhanced MRI after initial CT. A Markov cohort model simulated a cohort of patients after curative or adjuvant treatment, with follow-up over the remaining life expectancy. We analyzed mean life-years gain, quality-adjusted life-years (QALYs), costs per person, and incremental cost-effectiveness ratio (ICER). To evaluate results, we performed one-way, two-way, and probabilistic sensitivity analyses.The life expectancies and QALY were 7.22 years and 5.08 for the conventional CT strategy and 7.79 years and 5.52 for the gadoxetic acid-enhanced MRI strategy, respectively. The expected costs were $99,770 for conventional CT and $105,025 for gadoxetic acid-enhanced MRI in the United States. The ICER with gadoxetic acid-enhanced MRI was $11,957, as opposed to that with conventional CT, which was lower than the cost-effectiveness threshold of $50,000/QALY. One-way, two-way, and probabilistic sensitivity analyses showed unchanged results over an acceptable range.Gadoxetic acid-enhanced MRI after CT is cost-effective for detecting additional HCC in patients with early-stage HCC who can undergo curative treatment (besides liver transplantation). The cost-effectiveness of gadoxetic acid-enhanced MRI may be considered in the management of patients with early-stage HCC during staging.

We evaluated the cost-effectiveness of a gadoxetic acid-enhanced MRI (EOB-MRI) strategy compared with conventional MRI strategy and biopsy to differentiate focal nodular hyperplasia (FNH) from hepatocellular adenoma (HCA). A decision tree model was constructed to compare the cost-effectiveness of EOB-MRI, conventional MRI with extracellular contrast agents, and biopsy as the initial diagnostic modality in patients with incidentally detected focal liver lesions suspected of being FNH or HCA. We analysed the cost and effectiveness, i.e. probability of successful diagnosis of each strategy. Costs were based on utilisation rates and Medicare reimbursements in the USA and South Korea. In the base case analysis of our decision tree model, the effectiveness of the three strategies was similar. The cost of the EOB-MRI strategy ($1283 in USA, $813 in South Korea) was lowest compared with the biopsy strategy ($1725 in USA, $847 in South Korea) and the conventional MRI strategy ($1750 in USA, $962 in South Korea). One-way, two-way and probabilistic sensitivity analysis showed unchanged results over an acceptable range. EOB-MRI strategy is the most cost-effective strategy for differentiating FNH from HCA in patients with incidentally detected focal liver lesions in a non-cirrhotic liver. • The effectiveness of the three strategies was similar. • The cost of the EOB-MRI strategy was lowest. • EOB-MRI strategy is the most cost-effective for differentiating FNH from HCA.

A large volume of mobile data is being generated and shared among mobile devices such as smartphones. Most of the mobile platforms provide a user with a keyword-based full text search (FTS) in order to search for mobile data. However, FTS only returns the data corresponding to the keywords given by users as results without considering a user’s query intention. To overcome this limitation, we propose a semantically enhanced keyword-based search method. Although there are various semantic search techniques, it is hard to apply existing methods to mobile devices just as they are. This is caused by the characteristics of mobile devices such as isolated database structures and limited computing resources. To enable semantic search on mobile devices, we also propose a lightweight mobile ontology. Experimental results from the prototype implementation of the proposed method show that the proposed method provides a better user experience than the conventional FTS and returns accurate search results in an acceptable response time.

Fog computing is a promising paradigm in terms of extending cloud computing to an edge network. In a broad sense, fog computing in Internet-of-Vehicles(IoV) provides low-latency services since fog nodes are closely located with moving cars and are locally distributed. In this paper, we propose a fog computing architecture based on a publish/subscribe model. After that, we describe a traffic congestion control scenario using a smart traffic light system which operates on top of the proposed architecture. Furthermore, we propose an upper-level domain ontology in order to enhance the expressivity of knowledge and describe a variety of semantic properties which interlink spatial information in IoV. Finally, we present an active rule where supports the exchange of event-driven messages between publishing and subscribing fog nodes.

Background A randomized clinical trial has found that the addition of erlotinib to gemcitabine (GEM-E) for pancreatic cancer led to a modest increase in survival. The aim of this national population-based retrospective study was to compare the effectiveness of GEM-E to GEM alone for pancreatic cancer patients in real clinical practice. Methods Patients with pancreatic cancer (ICD-10: C25) with prescription claims of gemcitabine or erlotinib between Jan 1, 2007 and Dec 31, 2012 were retrospectively identified from the Korean Health Insurance claims database. To be included in the study population, patients were required to have had a histological or cytological diagnosis within one year before chemotherapy. Patients treated with prior radiotherapy, surgery, or chemotherapy were excluded to reduce heterogeneity. Overall survival from the initiation of therapy and the medical costs of GEM-E and GEM were compared. Results A total of 4,267 patients were included in the analysis. Overall survival was not significantly longer in patients treated with GEM-E (median 6.77 months for GEM-E vs. 6.68 months for GEM, p = 0.0977). There was also no significant difference in the respective one-year survival rates (27.0 % vs. 27.3 %; p = 0.5988). Multivariate analysis using age, gender, and comorbidities as covariates did not reveal any significant differences in survival. Based on this relative effectiveness, the incremental cost per life year gained over GEM was estimated at USD 70,843.64 for GEM-E. Conclusions GEM-E for pancreatic cancer is not more effective than GEM in a real-world setting, and it does not provide reasonable cost-effectiveness over GEM. Electronic supplementary material The online version of this article (doi:10.1186/s12885-016-2482-z) contains supplementary material, which is available to authorized users.

Hazard ration of survival. (DOCX 16 kb)

This study aimed to estimate Korean preference weights for EQ-5D-5L based on values elicited from Korean population by applying the EuroQol Valuation Technology (EQ-VT) program and the standard protocol by the EuroQol Group. The multistage quota sampling method was used to recruit 1085 subjects from the general population in Korea. Each respondent valuated 10 health states using the composite time trade-off (cTTO) and 7 health states using discrete choice experiment. The EQ-VT program was developed by the EuroQol Group and translated into Korean with the Korean research team. Computer-assisted, face-to-face interviews were conducted. A range of predictive models were explored using cTTO. The most appropriate model was determined after assessing goodness of fit, logical consistency, and parsimony. Of 3206 contacted, 1085 subjects completed interviews (33.8 %) and 1080 were used for modeling. A model with dummy variables for each level of severity and dimension and a term that picked up whether any dimension in the state was at level 4 or 5 was selected as the best predictive model. All coefficients of the final model were statistically significant and logically consistent. In addition, it was parsimonious. This model had mean absolute error of 0.027, and the absolute error for all 86 health states was below 0.1. The final model built in this study appears to predict the utilities of the states which were valuated directly. This model could be used to interpolate quality weights for all EQ-5D-5L health states.

Introduction:Early detection of primary hepatocellular carcinoma (PHC) patients with cirrhosis is critical to enhance PHC patients’ survival rates and to save medical costs. The study aimed to generate real world evidence to support the importance for early detection of PHC patients, and this evidence will contribute to a cost effectiveness analysis of the national liver cancer surveillance program.Methods:A retrospective analysis was performed on 98,275 PHC patients with cirrhosis in the National Center Cancer Registry from 2005 to 2014, linked to the Korea National Health Insurance claims database. The hazard ratio (HR) of mortality within five years and medical costs for the patients were compared by surveillance, epidemiology, and end results (SEER) stage.Results:There were differences in survival rates and medical costs depending on their characteristics including sex, age at diagnosis, SEER stage and types of initial treatment of cancer. The HR of mortality within five years of the PHC patients with distant stage versus local stage was 3.36 with 95% Confidence Interval (95% CI: 3.33–3.38) which is higher than those of the patients with regional stage (HR 1.93, 95% CI: 1.92–1.95). The estimated annual medical cost was USD 38,208 with standard deviation (SD) 54,399 for localized stage but USD 16,345 (SD 42,377) for distant stage.Conclusions:If PHC patients with cirrhosis were detected at early stage, their survival rates would be clinically better with a big saving for medical costs than if they were detected at distant stage. This result itself highlights that importance of the national liver cancer surveillance program. Future studies are indicated to apply these quantitative results into the cost-effectiveness analysis of the Korean national liver cancer surveillance program.

To highlight the societal burden of HBV infection in South Korea, we estimated the annual societal costs of HBV-related diseases for the year 2005. For the economic costs of HBV-infection-related diseases estimate, baseline data was collected from the Health Insurance Review and Assessment Service (HIRA) database. To complement data from the HIRA database, hospital charts from sample hospitals was reviewed and patient surveys were conducted. In 2005, the societal cost of HBV infection was 1.937 trillion KRW, including 474,642 million KRW of direct costs and 1.463 trillion KRW of indirect costs. The cost breakdown by disease was CHB at 465,167 million KRW, cirrhosis at 533,449 million KRW, hepatocellular carcinoma at 863,940 million KRW, and liver transplantation at 74,635 million KRW. The estimated amount is equivalent to 0.24% of the 2005 Korean GDP. This analysis emphasizes how important the prevention and treatment of these diseases are from the perspectives of the Korean society.

Objectives Depression and diabetes are closely biologically and behaviorally intertwined. We examined the impact of comorbid diabetes mellitus on the incidence of serious psychiatric outcomes among patients with depression. Methods We used claims data from the Korean Health Insurance Review & Assessment Service database of patients who were diagnosed with depression within one year of an index prescription for antidepressants between January 2007 and June 2008. We investigated the association between the comorbidity of diabetes mellitus and serious psychiatric outcomes of depression, such as psychiatric hospitalization, psychiatric emergency room visits, and suicide attempts. Results Among 200,936 patients with depression, 74,160 (36.9%) had diabetes mellitus, including 57,418 (28.6%) with complications. The incidence of serious psychiatric outcomes was 3.3% in patients with depression without diabetes and 6.7% in patients with depression and diabetes mellitus. Patients with depression and diabetes mellitus complications showed higher rates of serious outcomes than that did those without diabetes mellitus complications (odds ratio, 1.19; 95% confidence interval, 1.11–1.13). Similarly, depressed patients with micro and macrovascular diabetic complications were more likely to experience serious outcomes than those without diabetes mellitus complications (odds ratio, 2.2; 95% confidence interval, 2.07–2.34). Conclusions Our results showed that comorbid diabetes mellitus can increase the risk of serious outcomes of depression, such as suicide and hospitalization, and thus may alter the antidepressants prescription patterns and healthcare service use among patients with depressive disorders.

In this paper, we propose a method for processing spatiotemporal queries on semantic data streams generated from diverse sensors. On the Internet of Things (IoT) environment, the number of mobile sensors greatly increases and their locations are becoming more important. IoT services may not be fully supported when only considering the temporal feature of streaming data. Accordingly, stream processing should be performed with consideration into both temporal and spatial factors. However, existing researches have a limitation of processing spatial queries since they focus on the temporal processing of streaming data. To support spatiotemporal query processing on semantic data streams, we propose a query language, which integrates temporal and geospatial properties. Specifically, we construct a spatiotemporal index to process the proposed spatiotemporal query language efficiently. The experimental results with a prototype implementation show that the proposed method processes spatiotemporal queries in an acceptable time.

Background: The psychometric properties of the Korean Short Form-12 Health Survey, version 2 (SF-12 v2) have not been assessed in the general population. Therefore, the aim of our study was to evaluate the psychometric properties of the Korean version of the SF-12 v2 in the general population and to provide SF-12 v2 domain scores according to the general characteristics of the study population. Methods: A total of 1,000 participants from the general Korean population were recruited using a multistage quota sampling method. Psychometric properties were evaluated by descriptive statistics, validity, reliability, and exploratory factor analysis. Results: Item convergent and discriminant validity met the criteria established by the instrument developer. In the known-group comparison, male gender, age

Most of the mobile platforms provide a keyword based full text search (FTS) for users to find what they want. However, FTS has difficulties in dealing with the cases where a user cannot remember the exact keywords about target data or the number of search results is too many. To overcome these limitations of FTS, we propose a semantically enhanced method of searching for data on mobile devices along with mobile ontology. Experimental results of the proposed method show that our method provides accurate search results and is suitable for a mobile environment.

To apply semantic search to smartphones, we propose an efficient semantic search method based on a lightweight mobile ontology. Through a prototype implementation of a semantic search engine on an android smartphone, experimental results show that the proposed method provides more accurate search results and a better user experience compared to the conventional method.

This paper proposes a mobile search engine for smart devices, which effectively augments the result of local semantic search with useful Web information according to the intent and context of a mobile user. To support an intuitive query, we employ the conventional natural language user interface, which supports voice recognition. Through the prototype implementation of the proposed search engine, we find that it provides more meaningful search results semantically and contextually, compared with the conventional keyword-based search engines of mobile devices.

To enhance the incorrectness of keyword based search, we propose an efficient semantic search method based on a lightweight mobile ontology designed for smart mobile devices. In addition, we implement a prototype of semantic search engine working on Android smartphones and our prototype engine provides better user experience compared with keyword based search.

We shall show how to extract various important information on the motion of a target (enemy aircraft) from a sequence of its image frames obtained using a single imaging sensor which is mounted on an aircraft. Specifically, we present an algorithm which estimates following 12 parameters of an enemy aircraft from its image sequence; position (three), linear velocity (three), attitude (three), instantaneous angular velocity (three) and optional acceleration (three). To extract the attitude, we use a matching algorithm for the captured image frame to the modeled image. The objective function is the value of correlation between these two images, and we use simulated annealing and downhill simplex methods for maximizing the objective function. Finally, estimation of the position and the at titude is accomplished with a 12- state extended Kalman filter. Through simulations, we show that the proposed algorithm is superior to convention algorithms which do not use the attitude information.

We shall show how to extract various important information on the motion of a target (enemy aircraft) from a sequence of its image frames obtained from a single staring imaging sensor which is mounted on an aircraft. Specifically, we present an algorithm which estimates following 15 parameters of an enemy aircraft from its image sequence; position ( three) , linear velocity (three) , attitude (three) , instantaneous angular velocity (three) and acceleration (three) . The attitude information of the aircraft is obtained through the image matching operation. In simulations, we show thatthe proposed algorithm is superior to the conventional algorithms which do not use the attitude information.Keywords: imaging sensor, target tracking, Euler angles, Kalman filter 1. INTRODUCTION Over the past decade, many researchers have addressed the method of tracking with angle-only measurements from infrared (IR) sensors, foward-looking infrared (FUR) systems or CCD cameras to make its system less susceptible to

Purpose: The aim of this study was to determine the incremental effectiveness (the differences in progression-free survival between treatments), the incremental cost and the incremental cost-effectiveness of Genexol-PM compared to Paclitaxel when these drugs were used as treatment for patients with metastatic breast cancer. Methods: In the absence of any comparative direct evidence of the relative efficacy of Paclitaxel and Genexol-PM in this setting, a meta-analysis was conducted to determine the effects of the Paclitaxel on the health outcome. The decision tree model was constructed to evaluate the two treatment regimens. All the costs are in 2008 Korean Won (KW) and they were evaluated according to the 3rd party payer perspective, and the direct nonmedical and indirect costs were excluded. Results: When compared with Paclitaxel, Genexol-PM was shown to increase the response rate and the time to progression for patients with metastatic breast cancer. Although the overall treatment costs of Genexol-PM were slightly higher than those of Paclitaxel, Genexol-PM was associated with a delayed time to progression of 4.78 months per patient. The incremental cost effectiveness ratio for Genexol-PM versus Paclitaxel was KW 2,295,228 per year gained, which is far below the per capita GDP or the threshold of the willingness-to-pay in Korea. Conclusion: Compared with Paclitaxel, Genexol-PM for treating metastatic breast cancer is within the acceptable range of the cost-effectiveness ratio for medical intervention.

We appreciate Dr. Griffiths and colleagues' letter and their interest in our paper (1). They raised quite an important fundamental issue regarding the cost-benefit analysis (CBA) study. In CBA study, many assumptions are needed because of the lack of baseline data, which are the main weakness of this kind of study. We also did several assumptions, which are not verified in other studies. Dr. Griffiths et al. pointed out this. Therefore, the CBA study can be a reference in decision on health policy, especially when the policy on national immunization, but many other discussions are also very important. First, Dr. Griffiths et al. mentioned the unreasonably high price of the Hib vaccine in Korea in our study. It was one of most serious difficulties when gathering information regarding the vaccine price in the conduction of CBA. Thus far, the Hib vaccine has not been included in the Korean universal vaccination, and no data existed regarding the Hib vaccine. Therefore, we went through the vaccine prices in private clinics in 2003. The price of the Hib immunization is KRW 40,000, which includes the vaccine (around KRW27,000-30,000) and administration (around KRW10,000). Based on this data, it was assumed that there would be a 50% reduction in the administrative costs and a 20% reduction in the cost of the Hib vaccine. Therefore, we used KRW26,000 per shot. If the Hib immunization, however, was adopted in a national immunization schedule, the cost of the vaccine would probably be lower due to the economy of scale. Under another scenario, applying US$9 or US$11 per dose as in US, the benefit-cost ratio of Hib immunization in our paper (1) would be increased to 1.52-1.86, and then we could expect the economic efficiency to increase significantly. Second, Dr. Griffiths et al. pointed out the incidence data of Hib infection in this study. The study by Kim et al. (2), which was used to estimate the annual incidence in our study, was the first population-based study and the only report of annual incidence of Hib meningitis in Korea. However, this study did not include Hib infection apart from Hib meningitis. Unavoidably, we had to assume the incidence rate of Hib pneumonia based on United States' epidemiology data (3). To estimate Hib pneumonia in Korea, the proportion of meningitis and pneumonia in US was initially estimated, and the rate was then applied to estimate Korean Hib pneumonia incidence. In addition, the trend of Korean Hib meningitis is not equal to those of other Asian countries; therefore, it is also difficult to assume that Hib pneumonia is more prevalent than Hib meningitis, as in Indonesia. Although a sensitivity analysis was conducted to test uncertainty concerning the incidence of Hib infection, we also recognize that a problem of estimation of the incidence based on other country's epidemiologic data exists. We completely agree with Dr. Griffiths et al. last point. Economic evaluation is just one method available to provide information to help develop policy. Despite the low economic efficiency, we would like to point out that there are a significant number of cases of disease, deaths and sequelae that can be prevented by Hib vaccination. The scope of this study, however, was to test the efficiency of Hib universal immunization based on a CBA. Except for the CBA, other discussions regarding whether Hib vaccination should be included in the Korean national vaccination program are beyond the scope of this study; therefore, discussion on that point was not included.