Nikola Bogunovic

Feature Extraction for ECG Time-Series Mining Based on Chaos Theory

Chaos theory applied to ECG feature extraction is presented in this article. Several chaos methods, including phase space and attractors, correlation dimension, spatial filling index, central tendency measure and approximate entropy are explained in detail. A new feature extraction environment called ECG chaos extractor has been created in order to apply these chaos methods. System model and program functions are presented. Some of the obtained results are listed. Future work in this field of research is discussed.

A review of feature selection methods with applications

Feature selection (FS) methods can be used in data pre-processing to achieve efficient data reduction. This is useful for finding accurate data models. Since exhaustive search for optimal feature subset is infeasible in most cases, many search strategies have been proposed in literature. The usual applications of FS are in classification, clustering, and regression tasks. This review considers most of the commonly used FS techniques. Particular emphasis is on the application aspects. In addition to standard filter, wrapper, and embedded methods, we also provide insight into FS for recent hybrid approaches and other advanced topics.

A programming model for composing data-flow collaborative applications

Distributed systems are essential for many real world applications. The paper presents an experimental programming model that enables visual configuration, deployment and control of data flow based collaborative systems, a class of distributed applications (DA). The programming model solves the problem of interoperability among DA functional components through the introduction of a fast middleware network software layer and by implementing a transparent message based communication between processes executing on machines connected by a local area network. At the next higher level, the programming model allows deployment and interconnection of encapsulated modules by logical composition of the entire collaborative application. The project differs widely from the existing systems in the network communication overhead, and the user interface design.

Impact of censoring on learning Bayesian networks in survival modelling

OBJECTIVE Bayesian networks are commonly used for presenting uncertainty and covariate interactions in an easily interpretable way. Because of their efficient inference and ability to represent causal relationships, they are an excellent choice for medical decision support systems in diagnosis, treatment, and prognosis. Although good procedures for learning Bayesian networks from data have been defined, their performance in learning from censored survival data has not been widely studied. In this paper, we explore how to use these procedures to learn about possible interactions between prognostic factors and their influence on the variate of interest. We study how censoring affects the probability of learning correct Bayesian network structures. Additionally, we analyse the potential usefulness of the learnt models for predicting the time-independent probability of an event of interest. METHODS AND MATERIALS We analysed the influence of censoring with a simulation on synthetic data sampled from randomly generated Bayesian networks. We used two well-known methods for learning Bayesian networks from data: a constraint-based method and a score-based method. We compared the performance of each method under different levels of censoring to those of the naive Bayes classifier and the proportional hazards model. We did additional experiments on several datasets from real-world medical domains. The machine-learning methods treated censored cases in the data as event-free. RESULTS We report and compare results for several commonly used model evaluation metrics. On average, the proportional hazards method outperformed other methods in most censoring setups. As part of the simulation study, we also analysed structural similarities of the learnt networks. Heavy censoring, as opposed to no censoring, produces up to a 5% surplus and up to 10% missing total arcs. It also produces up to 50% missing arcs that should originally be connected to the variate of interest. CONCLUSION Presented methods for learning Bayesian networks from data can be used to learn from censored survival data in the presence of light censoring (up to 20%) by treating censored cases as event-free. Given intermediate or heavy censoring, the learnt models become tuned to the majority class and would thus require a different approach.

HRVFrame: Java-based framework for feature extraction from cardiac rhythm

Heart rate variability (HRV) analysis can be successfully applied to automatic classification of cardiac rhythm abnormalities. This paper presents a novel Java-based computer framework for feature extraction from cardiac rhythms. The framework called HRVFrame implements more than 30 HRV linear time domain, frequency domain, time-frequency domain, and nonlinear features. Output of the framework in the form of .arff files enables easier medical knowledge discovery via platforms such as RapidMiner or Weka. The scope of the framework facilitates comparison of models for different cardiac disorders. Some of the features implemented in the framework can also be applied to other biomedical time-series. The thorough approach to feature extraction pursued in this work is also encouraged for other types of biomedical time-series.

Time series classification based on qualitative space fragmentation

In knowledge discovery and data mining from time series the goal is to detect interesting patterns in the series that may help a human to better recognize the regularities in the observed variables and thereby improve the understanding of the system. Ideally, knowledge discovery algorithms use time series representations that are close to those that are used by a human. The impressive pattern recognition capabilities of the human brain help to establish connections between different time series or different parts of a single time series on the basis of their visual appearance. When dealing with time series data there are two main objectives: (i) prediction of future behavior based on past behaviors and (ii) description (explanation) of time series data. Description of time series data can be used for generalization, clustering and classification. In this paper, a novel time series classification method based on Qualitative Space Fragmentation is presented. The main characteristics of the presented method are expansion and coding of quantitative time series data together with extraction of symbolic and numeric features based on human visual perception. The expansion and coding process results in the creation of a qualitative difference vector. The qualitative difference vector conveys full information on the variation of the particular time series and can be seen as a single point in m-dimensional qualitative-space. Symbolic and numeric features based on human visual perception are extracted from the qualitative space and used for the decision tree construction that is later employed in time series classification. The application of the proposed method is demonstrated through two different case studies. In the first case study, the method was tested in the context of synthetic Control Chart Pattern data, which are time series developed for the assessment of the statistical process control. The obtained results were compared with the standard Qualitative Similarity Index method. In the second case study the method was tested in the field of analytic chemistry - polarography, an electrochemical method for analyzing solutions containing reducible or oxidizable substances.

Decision Tree Ensembles in Biomedical Time-Series Classification

There are numerous classification methods developed in the field of machine learning. Some of these methods, such as artificial neural networks and support vector machines, are used extensively in biomedical time-series classification. Other methods have been used less often for no apparent reason. The aim of this work is to examine the applicability of decision tree ensembles as strong and practical classification algorithms in biomedical domain. We consider four common decision tree ensembles: AdaBoost.M1+C4.5, Multi- Boost+C4.5, random forest, and rotation forest. The decision tree ensembles are compared with SMO-based support vector machines classifiers (linear, squared polynomial, and radial kernel) on three distinct biomedical time-series datasets. For evaluation purposes, 10x10-fold cross-validation is used and the classifiers are measured in terms of sensitivity, specificity, and speed of model construction. The classifiers are compared in terms of statistically significant winslosses-ties on the three datasets. We show that the overall results favor decision tree ensembles over SMO-based support vector machines. Preliminary results suggest that AdaBoost.M1 and MultiBoost are the best of the examined classifiers, with no statistically significant difference between them. These results should encourage the use of decision tree ensembles in biomedical time-series datasets where optimal model accuracy is sought.

Random Forest-Based Classification of Heart Rate Variability Signals by Using Combinations of Linear and Nonlinear Features

The goal of this paper is to assess various combinations of heart rate variability (HRV) features in successful classification of four different cardiac rhythms. The rhythms include: normal, congestive heart failure, supraventricular arrhythmia, and any arrhythmia. We approach the problem of automatic cardiac rhythm classification from HRV by employing several features’ schemes. The schemes are evaluated using the random forest classifier. We extracted a total of 78 linear and nonlinear features. Highest results were achieved for normal/supraventricular arrhythmia classification (93%). A feature scheme consisting of: time domain (SDNN, RMSSD, pNN20, pNN50, HTI), frequency domain (Total PSD, VLF, LF, HF, LF/HF), SD1/SD2 ratio, Fano factor, and Allan factor features demonstrated very high classification accuracy, comparable to the results for all extracted features. Results show that nonlinear features have only minor influence on overall classification accuracy.

Verification of mutual exclusion algorithms with SMV system

A mutual exclusion algorithm can exhibit intricate behavior for which correctness can be hard to establish. We demonstrate automatic verification of five algorithms by symbolic model checking. We used the SMV tool which enables property specification in computation tree logic and allows us to impose fairness constraints on a model. For each algorithm we verify safety, liveness, nonblocking and no strict ordering properties.

STIMONT: a core ontology for multimedia stimuli description

Affective multimedia documents such as images, sounds or videos elicit emotional responses in exposed human subjects. These stimuli are stored in affective multimedia databases and successfully used for a wide variety of research in psychology and neuroscience in areas related to attention and emotion processing. Although important all affective multimedia databases have numerous deficiencies which impair their applicability. These problems, which are brought forward in the paper, result in low recall and precision of multimedia stimuli retrieval which makes creating emotion elicitation procedures difficult and labor-intensive. To address these issues a new core ontology STIMONT is introduced. The STIMONT is written in OWL-DL formalism and extends W3C EmotionML format with an expressive and formal representation of affective concepts, high-level semantics, stimuli document metadata and the elicited physiology. The advantages of ontology in description of affective multimedia stimuli are demonstrated in a document retrieval experiment and compared against contemporary keyword-based querying methods. Also, a software tool Intelligent Stimulus Generator for retrieval of affective multimedia and construction of stimuli sequences is presented.