Goran Krstačić

Active subgroup mining: a case study in coronary heart disease risk group detection

This paper presents an approach to active mining of patient records aimed at discovering patient groups at high risk for coronary heart disease (CHD). The approach proposes active expert involvement in the following steps of the knowledge discovery process: data gathering, cleaning and transformation, subgroup discovery, statistical characterization of induced subgroups, their interpretation, and the evaluation of results. As in the discovery and characterization of risk subgroups, the main risk factors are made explicit, the proposed methodology has high potential for patient screening and early detection of patient groups at risk for CHD.

Heart rate variability and nonlinear dynamic analysis in patients with stress-induced cardiomyopathy

Complexity-based analyses may quantify abnormalities in heart rate variability (HRV). The aim of this study was to investigate the clinical and prognostic significances of dynamic HRV changes in patients with stress-induced cardiomyopathy Takotsubo syndrome (TS) by means of linear and nonlinear analysis. Patients with TS were included in study after complete noninvasive and invasive cardiovascular diagnostic evaluation and compared to an age and gender matched control group of healthy subjects. Series of R–R interval and of ST–T interval values were obtained from 24-h ECG recordings after digital sampling. HRV analysis was performed by ‘range rescaled analysis’ to determine the Hurst exponent, by detrended fluctuation analysis to quantify fractal long-range correlation properties, and by approximate entropy to assess time-series predictability. Short- and long-term fractal-scaling exponents were significantly higher in patients with TS in acute phases, opposite to lower approximate entropy and Hurst exponent, but all variables normalized in a few weeks. Dynamic HRV analysis allows assessing changes in complexity features of HRV in TS patients during the acute stage, and to monitor recovery after treatment, thus complementing traditional ECG and clinically analysis.

Non-linear analysis of heart rate variability in patients with coronary heart disease

The article emphasizes clinical and prognostic significance of non-linear measures of the heart rate variability, applied on the group of patients with coronary heart disease (CHD) and age-matched healthy control group. Three different methods were applied: Hurst exponent (H), Detrended Fluctuation Analysis (DFA) and approximate entropy (ApEn). Hurst exponent of the R-R series was determined by the range rescaled analysis technique. DFA was used to quantify fractal long-range-correlation properties of heart rate variability. Approximate entropy measures the unpredictability of fluctuations in a time series. It was found that the short-term fractal scaling exponent (/spl alpha//sub 1/) is significantly lower in patients with CHD (0.93 /spl plusmn/ 0.07 vs. 1.09 /spl plusmn/ 0.04; p < 0.001). The patients with CHD had lower Hurst exponent in each program of exercise test separately, as well as approximate entropy than healthy control group (P < 0.001).

Inconsistency Tests for Patient Records in a Coronary Heart Disease Database

The work presents the results of inconsistency detection experiments on the data records of an atherosclerotic coronary heart disease database collected in the regular medical practice. Medical expert evaluation of some preliminary inductive learning results have demonstrated that explicit detection of outliers can be useful for maintaining the data quality of medical records and that it might be a key for the improvement of medical decisions and their reliability in the regular medical practice. With the intention of on-line detection of possible data inconsistences, sets of confirmation rules have been developed for the database and their test results are reported in this work.

Clinical data analysis based on iterative subgroup discovery: experiments in brain ischaemia data analysis

Abstract This paper presents a case study of the process of insightful analysis of clinical data collected in regular hospital practice. The approach is applied to a database describing patients suffering from brain ischaemia, either permanent as brain stroke with positive computer tomography (CT) or reversible ischaemia with normal brain CT test. The goal of the analysis is the extraction of useful knowledge that can help in diagnosis, prevention and better understanding of the vascular brain disease. This paper demonstrates the applicability of subgroup discovery for insightful data analysis and describes the expert’s process of converting the induced rules into useful medical knowledge. Detection of coexisting risk factors, selection of relevant discriminative points for numerical descriptors, as well as the detection and description of characteristic patient subpopulations are important results of the analysis. Graphical representation is extensively used to illustrate the detected dependencies in the available clinical data.

Some important R-R interval based paroxysmal atrial fibrillation predictors

A trial fibrillation is the most common sustained cardiac arrhythmia. The result of series of machine learning experiments is detection of some promising paroxysmal atrial fibrillation predictors. Based on ratio of short and long R-R intervals there is a possibility to generate rules for PAF screening and predicting. For PAF screening the calculated ratio were 2.00 for successive R-R intervals. The problem of imminent PAF prediction is much more difficult and the concept of normalisation hail to be implemented. The optimal seems to be ratio between the shortest and the longest R-R interval, which was at least 1.75 times larger than ratio during the normalisation time for the same patient. Also it was detected that maximal distance of the longest and the shortest R-R intervals should be up to six R-R intervals.

The Chaos Theory and Non-linear Dynamics in Heart Rate Variability in Patients with Heart Failure

This study evaluate and quantify the non-linear dynamic changes of heart rate variability based on ldquochaos theoryrdquo and fractal mathematics in 250 patients with heart failure during 12 months. Some different non-linear methods were applied: fractal dimension (FD), detrented fluctuation analysis (DFA) and approximate entropy (ApEn). Fractal correlation properties and fractal dimension in this study may reflect altered neuroanatomic interaction that may predispose to the development of severe HF. It was found that the short-term fractal scaling exponent (alpha1) is significantly lower in patients with HF. The patients with HF had also lower approximate entropy and higher fractal dimension with positive impact of modern HE therapy.

Coronary Heart Disease Patient Models Based on Inductive Machine Learning

The work presents a model construction process which is a combination of the inductive learning based detection of interesting subgroups, comparative statistical analyses of risk factors for these groups, and expert knowledge interpretation of the results. The induced models describe population subgroups with unproportionately high rate of the disease what might be helpful in the prevention process.

Optimizing the detection of characteristic waves in ECG based on exploration of processing steps combinations

In this paper, algorithms for detection of characteristic waves in ECG are examined and modified. We distinguished four processing steps of detection algorithms: noise and artefacts reduction, transformations, fiducial marks selection of wave candidates, and decision rule. Several algorithms for detection of QRS, P, and T waves are explored through combinations of processing steps, in order to achieve accurate detection results. Algorithms are tested on public available ECG databases with both QRS and P and T waves annotations. We found that, depending on the database, the combination of Sun Yan’s MMF or MMD methods with Elgendi’s algorithm works best for QRS detection (Se = 99.77% +P = 99.72% for MMF on MIT-BIH Arrhythmia Database and Se = 99.90% +P = 99.89% for MMD on QT Database), while P and T waves were best detected using only Elgendi’s algorithm (P waves: Se = 60.84% +P = 59.61%, T waves: Se = 88.79% +P = 93.55% on MIT-BIH Arrhythmia Database). Our work shows that combining the best proposed methods in literature may lead to improvements in ECG waves detection, although P and T waves detection is still less than satisfactory and warrants further research.

Correlation between Concentration of Air Pollutants and Occurrence of Cardiac Arrhythmias in a Region with Humid Continental Climate

In this study, we investigated the correlation of air temperature, pressure and concentration of air pollutants with the rate of admissions for cardiac arrhythmias at two clinical centers in the area with a humid continental climate. This retrospective study included 3749 patients with arrhythmias admitted to emergency department (ED). They were classified into four groups: supraventricular tachycardia (SVT), ventricular tachycardia (VT), atrial fibrillation/undulation (Afib/Aund), and palpitations (with no ECG changes, or with sinus tachycardia and extrasystoles). The number of patients, values of meteorological parameters (average daily values of air temperature, pressure and relative humidity) and concentrations of air pollutants (particles of dimensions ~10 micrometers or less (PM(10)), ozone (O(3)) and nitrogen dioxide (NO(2))) were collected during a two-year period ( July 2008-June 2010). There were 1650 (44.0%), 1525 (40.7%), 451 (12.0%) and 123 (3.3%) patients with palpitations, Afib/Aund, SVT and VT, respectively. Spearman’s correlation yielded positive correlation between the occurrence of arrhythmias and air humidity on the day (r=0.07), and 1 (r=0.08), 2 (r=0.09) and 3 days before (r=0.09), and NO(2) particles on the day (r=0.08) of ED admission; palpitations and air humidity on the day (r=0.11), and 1 (r=0.09), 2 (r=0.07) and 3 days before (r=0.10), and PM(10) (r=0.11) and NO(2) (r=0.08) particles on the day of ED admission; and Afi b/Aund and air humidity 2 days before (r=0.08) ED admission (p<0.05 all). In conclusion, there was a very weak positive correlation of the occurrence of cardiac arrhythmias with air humidity and concentration of air pollutants in the region with a humid continental climate.