Ljiljana Majnarić

Visual analytics for concept exploration in subspaces of patient groups

Medical doctors and researchers in bio-medicine are increasingly confronted with complex patient data, posing new and difficult analysis challenges. These data are often comprising high-dimensional descriptions of patient conditions and measurements on the success of certain therapies. An important analysis question in such data is to compare and correlate patient conditions and therapy results along with combinations of dimensions. As the number of dimensions is often very large, one needs to map them to a smaller number of relevant dimensions to be more amenable for expert analysis. This is because irrelevant, redundant, and conflicting dimensions can negatively affect effectiveness and efficiency of the analytic process (the so-called curse of dimensionality). However, the possible mappings from high- to low-dimensional spaces are ambiguous. For example, the similarity between patients may change by considering different combinations of relevant dimensions (subspaces). We demonstrate the potential of subspace analysis for the interpretation of high-dimensional medical data. Specifically, we present SubVIS, an interactive tool to visually explore subspace clusters from different perspectives, introduce a novel analysis workflow, and discuss future directions for high-dimensional (medical) data analysis and its visual exploration. We apply the presented workflow to a real-world dataset from the medical domain and show its usefulness with a domain expert evaluation.

Knowledge discovery of drug data on the example of adverse reaction prediction

BackgroundAntibiotics are the widely prescribed drugs for children and most likely to be related with adverse reactions. Record on adverse reactions and allergies from antibiotics considerably affect the prescription choices. We consider this a biomedical decision-making problem and explore hidden knowledge in survey results on data extracted from a big data pool of health records of children, from the Health Center of Osijek, Eastern Croatia.ResultsWe applied and evaluated a k-means algorithm to the dataset to generate some clusters which have similar features. Our results highlight that some type of antibiotics form different clusters, which insight is most helpful for the clinician to support better decision-making.ConclusionsMedical professionals can investigate the clusters which our study revealed, thus gaining useful knowledge and insight into this data for their clinical studies.

An adaptive annotation approach for biomedical entity and relation recognition

Abstract In this article, we demonstrate the impact of interactive machine learning: we develop biomedical entity recognition dataset using a human-into-the-loop approach. In contrary to classical machine learning, human-in-the-loop approaches do not operate on predefined training or test sets, but assume that human input regarding system improvement is supplied iteratively. Here, during annotation, a machine learning model is built on previous annotations and used to propose labels for subsequent annotation. To demonstrate that such interactive and iterative annotation speeds up the development of quality dataset annotation, we conduct three experiments. In the first experiment, we carry out an iterative annotation experimental simulation and show that only a handful of medical abstracts need to be annotated to produce suggestions that increase annotation speed. In the second experiment, clinical doctors have conducted a case study in annotating medical terms documents relevant for their research. The third experiment explores the annotation of semantic relations with relation instance learning across documents. The experiments validate our method qualitatively and quantitatively, and give rise to a more personalized, responsive information extraction technology.

Analysis of Patient Groups and Immunization Results Based on Subspace Clustering

Biomedical experts are increasingly confronted with what is often called Big Data, an important subclass of high-dimensional data. High-dimensional data analysis can be helpful in finding relationships between records and dimensions. However, due to data complexity, experts are decreasingly capable of dealing with increasingly complex data. Mapping higher dimensional data to a smaller number of relevant dimensions is a big challenge due to the curse of dimensionality. Irrelevant, redundant, and conflicting dimensions affect the effectiveness and efficiency of analysis. Furthermore, the possible mappings from high- to low-dimensional spaces are ambiguous. For example, the similarity between patients may change by considering different combinations of relevant dimensions (subspaces). We show the potential of subspace analysis for the interpretation of high-dimensional medical data. Specifically, we analyze relationships between patients, sets of patient attributes, and outcomes of a vaccination treatment by means of a subspace clustering approach. We present an analysis workflow and discuss future directions for high-dimensional (medical) data analysis and visual exploration.

Interactive and Iterative Annotation for Biomedical Entity Recognition

In this paper, we demonstrate the impact of interactive machine learning for the development of a biomedical entity recognition dataset using a human-into-the-loop approach: during annotation, a machine learning model is built on previous annotations and used to propose labels for subsequent annotation. To demonstrate that such interactive and iterative annotation speeds up the development of quality dataset annotation, we conduct two experiments. In the first experiment, we carry out an iterative annotation experimental simulation and show that only a handful of medical abstracts need to be annotated to produce suggestions that increase annotation speed. In the second experiment, clinical doctors have conducted a case study in annotating medical terms documents relevant for their research. The experiments validate our method qualitatively and quantitatively, and give rise to a more personalized, responsive information extraction technology.

On Patient’s Characteristics Extraction for Metabolic Syndrome Diagnosis: Predictive Modelling Based on Machine Learning

The work presented in this paper demonstrates how different data mining approaches can be applied to extend conventional combinations of variables determining the Metabolic Syndrome with new influential variables, which are easily available in the everyday physician‘s practice. The results have important consequences: patients with the Metabolic Syndrome can be recognized by using only some, one, or none of the conventional variables, when replaced with some other surrogate variables, available in patient health records, making diagnosis feasible in different work environments and at different time points of patient care. In addition, the results showed that there is a large diversity of patient groups, much larger than it was supposed earlier on when their identification was based on the conventional variables approach, indicating the underlying complexity of this syndrome. Finally, the discovered novel variables, indicating yet unknown pathogenetic pathways can be used to inspire future research.

A retrospective, case-control study on traditional environmental risk factors in inflammatory bowel disease in Vukovar-Srijem County, north-eastern Croatia, 2010

SummaryBackgroundTraditional environmental risk factors in inflammatory bowel disease (IBD), ulcerative colitis (UC) and Crohn’s disease (CD), were examined as part of the retrospective epidemiologic study conducted in Vukovar-Srijem County, north-eastern Croatia in 2010. The geographical variations in the frequency of IBD in Croatia have been observed, which is also the trend in the Central Eastern European region and Europe as a whole, indicating the influence of environmental and lifestyle factors. However, the data on the spread of environmental IBD risk factors are still limited. The purpose of this study was to analyse the traditional environmental risk factors in IBD on our cohort sample, including measles virus infection and vaccination (MMR vaccine—Mumps, Measles, Rubella), tonsillectomy, appendectomy, current and former cigarette smoking and use of oral contraceptives in women.MethodsThis retrospective, case-control study was performed as a part of a wider epidemiologic study aimed at assessing the incidence, prevalence and clinical expression of IBD, in Vukovar-Srijem County (population: 204,768; 2001), which is a lesser developed part of the continental Croatia that experienced deep demographic changes in the recent past. IBD patients were identified according to the hospital’s patient records. There were 119 UC patients and 31 CD patients of a total of 150 patients in the cohort. A total of 150 individuals, volunteers, not having a diagnosis of IBD, age- and sex-matched, were used as the control group. Information on examined risk factors were obtained from all subjects in a previously conducted interview. Patients were contacted personally or by phone and interviewed by a gastroenterologist.ResultsThere were no differences in the number of smokers, former smokers and non-smokers, between UC and CD patients and the controls, nor in the duration of smoking (years), in current smokers and ex-smokers. Only marginally significant longer time of non-smoking, in ex-smokers was found in IBD patients, compared to the controls, more pronounced in CD patients (p = 0.05). No difference was found in relation to tonsillectomy and risk of IBD. There was no difference in the number of female IBD patients and women from the controls using oral contraceptives. Duration (years) of oral contraceptives use was longer in women from the controls than in female IBD patients (p < 0.001). Frequency of appendectomy was the lowest in UC patients, compared to the controls and CD patients (3.4, 12.0 and 38.7 %, respectively) (p < 0.001). No difference was found in relation to measles virus infection and risk of IBD. MMR vaccination rates were higher in CD patients (90.3 %), compared to UC patients and the controls (74.8 and 67.3 %, respectively) (p = 0.026).ConclusionsNo association was found between smoking and tonsillectomy and risk of IBD. Our results do not support the idea of oral contraceptives use as a risk factor for IBD. Frequency of appendectomy was the lowest in UC patients, suggesting that appendectomy decreases the risk of UC, contributing the earlier results. MMR vaccination seemed to be associated with Crohn’s disease. These results can add value to our understanding of the increasing incidence of IBD in Croatia and other Central Eastern European countries and can be introductory to future large-scale research.

Simple understandable analysis of medical data to support the diagnostic process

Medical diagnostic is a complex process consisting of many input variables, which the general practitioner (GP) or specialist should take into account before confirm the expected diagnosis. In the case of electronic records, they have an opportunity to support this process within simple understandable results of the correctly applied suitable methods from machine learning or statistics. We used a small sample of patients data from Croatia for experimental evaluation of this potential. We applied the methods as Welchs t-test, Pearson chi-square independence test, Youdens index, decision trees and simple K-Means. The cooperating medical expert evaluated the obtained results and confirmed the expected potential for daily medical practice.

Discovery of Characteristics of Patients with Increased Level ofInflammation

This paper is a study on knowledge discovery for the prediction of characteristics of older patients with increased level of inflammation. The etiology of inflammation is thought to be multifactorial and associated with the development of chronic aging diseases. Chronic low grade inflammation, expressed by slightly elevated serum concentrations of the inflammatory marker C-reactive protein (CRP), has been showed to be associated with increased frailty and overall and specific cardiovascular and noncardiovascular mortality. However, it has not been discovered before which conditions, taken all at once, are associated with elevated systems level of inflammation. To answer this question, we used the group of older primary health care attenders, burdened with multiple chronic conditions and described their health status by many aspects. The dataset was composed of 61 low-cost health parameters, many of which were data from patient health records. To predict the characteristics of patients with increased level of inflammation, we used a Linear Regression model and compared the results with some classfication algorithms. In this way, we selected 11 relevant predictors of inflammation and explained their meaning according to the existing knowledge. We could realise that many of them represent the components of the highly conserved functional network in which inflammation is the intermediate mechanism, linking these components together. These components include the metabolic, the neuroendocrine and the immune system, proposed as influencing each other during the development of age-related chronic diseases. We have also identified some new components, represented by the parameters indicating inflammation-mediated locomotor system disorders and the pituitary hormone prolactin serum concentration variations. This model, resulted from the knowledge discovery procedure, can be used to provide guidelines for further research on chronic low grade inflammation and for more practical purposes, to help physicians recognizing older persons who are at increased risk for frailty and death.

Nodular lymphoid hyperplasia complicated with ileal Burkitt's lymphoma in an adult patient with selective IgA deficiency

Highlights • Burkitts lymphoma in adults can occur on the basis of nodular lymphoid hyperplasia.• Nodular lymphoid hyperplasia is associated with selective IgA deficiency.• Nodular lymphoid hyperplasia associates IgA deficiency with Burkitts lymphoma.