PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals.

The newly inaugurated Research Resource for Complex Physiologic Signals, which was created under the auspices of the National Center for Research Resources of the National Institutes of Health, is intended to stimulate current research and new investigations in the study of cardiovascular and other complex biomedical signals. The resource has 3 interdependent components. PhysioBank is a large and growing archive of well-characterized digital recordings of physiological signals and related data for use by the biomedical research community. It currently includes databases of multiparameter cardiopulmonary, neural, and other biomedical signals from healthy subjects and from patients with a variety of conditions with major public health implications, including life-threatening arrhythmias, congestive heart failure, sleep apnea, neurological disorders, and aging. PhysioToolkit is a library of open-source software for physiological signal processing and analysis, the detection of physiologically significant events using both classic techniques and novel methods based on statistical physics and nonlinear dynamics, the interactive display and characterization of signals, the creation of new databases, the simulation of physiological and other signals, the quantitative evaluation and comparison of analysis methods, and the analysis of nonstationary processes. PhysioNet is an on-line forum for the dissemination and exchange of recorded biomedical signals and open-source software for analyzing them. It provides facilities for the cooperative analysis of data and the evaluation of proposed new algorithms. In addition to providing free electronic access to PhysioBank data and PhysioToolkit software via the World Wide Web (http://www.physionet. org), PhysioNet offers services and training via on-line tutorials to assist users with varying levels of expertise.

Power spectral density of unevenly sampled data by least-square analysis: performance and application to heart rate signals.

This work studies the frequency behavior of a least-square method to estimate the power spectral density of unevenly sampled signals. When the uneven sampling can be modeled as uniform sampling plus a stationary random deviation, this spectrum results in a periodic repetition of the original continuous time spectrum at the mean Nyquist frequency, with a low-pass effect affecting upper frequency bands that depends on the sampling dispersion. If the dispersion is small compared with the mean sampling period, the estimation at the base band is unbiased with practically no dispersion. When uneven sampling is modeled by a deterministic sinusoidal variation respect to the uniform sampling the obtained results are in agreement with those obtained for small random deviation. This approximation is usually well satisfied in signals like heart rate (HR) series. The theoretically predicted performance has been tested and corroborated with simulated and real HR signals. The Lomb method has been compared with the classical power spectral density (PSD) estimators that include resampling to get uniform sampling. We have found that the Lomb method avoids the major problem of classical methods: the low-pass effect of the resampling. Also only frequencies up to the mean Nyquist frequency should be considered (lower than 0.5 Hz if the HR is lower than 60 bpm). We conclude that for PSD estimation of unevenly sampled signals the Lomb method is more suitable than fast Fourier transform or autoregressive estimate with linear or cubic interpolation. In extreme situations (low-HR or high-frequency components) the Lomb estimate still introduces high-frequency contamination that suggest further studies of superior performance interpolators. In the case of HR signals we have also marked the convenience of selecting a stationary heart rate period to carry out a heart rate variability analysis.

Reduction of false arterial blood pressure alarms using signal quality assessment and relationships between the electrocardiogram and arterial blood pressure.

The paper presents an algorithm for reducing false alarms related to changes in arterial blood pressure (ABP) in intensive care unit (ICU) monitoring. The algorithm assesses the ABP signal quality, analyses the relationship between the electrocardiogram and ABP using a fuzzy logic approach and post-processes (accepts or rejects) ABP alarms produced by a commercial monitor. The algorithm was developed and evaluated using unrelated sets of data from the MIMIC database. By rejecting 98.2% (159 of 162) of the false ABP alarms produced by the monitor using the test set of data, the algorithm was able to reduce the false ABP alarm rate from 26.8% to 0.5% of ABP alarms, while accepting 99.8% (441 of 442) of true ABP alarms. The results show that the algorithm is effective and practical, and its use in future patient monitoring systems is feasible.

Analysis of the ST-T complex of the electrocardiogram using the Karhunen--Loève transform: adaptive monitoring and alternans detection.

The Karhunen-Loève transform (KLT) is applied to study the ventricular repolarisation period as reflected in the ST-T complex of the surface ECG. The KLT coefficients provide a sensitive means of quantitating ST-T shapes. A training set of ST-T complexes is used to derive a set of KLT basis vectors that permits representation of 90% of the signal energy using four KLT coefficients. As a truncated KLT expansion tends to favor representation of the signal over any additive noise, a time series of KLT coefficients obtained from successive ST-T complexes is better suited for representation of both medium-term variations (such as ischemic changes) and short-term variations (such as ST-T alternans) than discrete parameters such as the ST level or other local indices. For analysis of ischemic changes, an adaptive filter is described that can be used to estimate the KLT coefficient, yielding an increase in the signal-to-noise ratio of 10 dB (u = 0.1), with a convergence time of about three beats. A beat spectrum of the unfiltered KLT coefficient series is used for detection of ST-T alterans. These methods are illustrated with examples from the European ST-T Database. About 20% of records revealed quasi-periodic salvos of ischemic ST-T change episodes and another 20% exhibit repetitive, but not clearly periodic patterns of ST-T change episodes. About 5% of ischemic episodes were associated with ST-T alterans.

Long-term ST database: a reference for the development and evaluation of automated ischaemia detectors and for the study of the dynamics of myocardial ischaemia.

The long-term ST database is the result of a multinational research effort. The goal was to develop a challenging and realistic research resource for development and evaluation of automated systems to detect transient ST segment changes in electrocardiograms and for supporting basic research into the mechanisms and dynamics of transient myocardial ischaemia. Twenty-four hour ambulatory ECG records were selected from routine clinical practice settings in the USA and Europe, between 1994 and 2000, on the basis of occurrence of ischaemic and non-ischaemic ST segment changes. Human expert annotators used newly developed annotation protocols and a specially developed interactive graphic editor tool (SEMIA) that supported paperless editing of annotations and facilitated international co-operation via the Internet. The database contains 86 two- and three-channel 24 h annotated ambulatory records from 80 patients and is stored on DVD-ROMs. The database annotation files contain ST segment annotations of transient ischaemic (1155) and heart-rate related ST episodes and annotations of non-ischaemic ST segment events related to postural changes and conduction abnormalities. The database is intended to complement the European Society of Cardiology ST-T database and the MIT-BIH and AHA arrhythmia databases. It provides a comprehensive representation of 'real-world' data, with numerous examples of transient ischaemic and non-ischaemic ST segment changes, arrhythmias, conduction abnormalities, axis shifts, noise and artifacts.

Robust parameter extraction for decision support using multimodal intensive care data.

Digital information flow within the intensive care unit (ICU) continues to grow, with advances in technology and computational biology. Recent developments in the integration and archiving of these data have resulted in new opportunities for data analysis and clinical feedback. New problems associated with ICU databases have also arisen. ICU data are high-dimensional, often sparse, asynchronous and irregularly sampled, as well as being non-stationary, noisy and subject to frequent exogenous perturbations by clinical staff. Relationships between different physiological parameters are usually nonlinear (except within restricted ranges), and the equipment used to measure the observables is often inherently error-prone and biased. The prior probabilities associated with an individual's genetics, pre-existing conditions, lifestyle and ongoing medical treatment all affect prediction and classification accuracy. In this paper, we describe some of the key problems and associated methods that hold promise for robust parameter extraction and data fusion for use in clinical decision support in the ICU.

Development of the polysomnographic database on CD-ROM.

We have developed a polysomnographic database on CD-ROM. The data were obtained from 16 subjects with sleep apnea syndrome. The physiological signals include electroencephalogram, electromyogram, electrooculogram, invasive blood pressure, respiratory wave, oxygen saturation, and cardiac volume as measured by VEST method. The CD-ROM also include programs to analyze polysomnography (PSG) data. The CD-ROM has values: (i) for researchers investigating clinical physiology or non-linear dynamics during sleep apnea syndrome; (ii) for engineers developing a new algorithm for the computerized analysis of PSG data related to sleep apnea syndrome; (iii) for students learning sleep physiology.

PhysioNet: a research resource for studies of complex physiologic and biomedical signals.

PhysioNet (http://www.physionet.org/) is a web-based resource supplying well-characterized physiologic signals and related open-source software to the biomedical research community. Inaugurated in September 1999 under the auspices of the NIH's National Center for Research Resources (NCRR), PhysioNet provides an on-line forum for free dissemination and exchange of research data and software, with facilities for cooperative analysis of data and evaluation of new analytic methods. As of September 2000, PhysioBank, the data archive made available via PhysioNet, contained roughly 35 gigabytes of recorded signals and annotations. PhysioNet is a public service of the Research Resource for Complex Physiologic Signals, a cooperative project initiated by researchers at Boston's Beth Israel Deaconess Medical Center/Harvard Medical School, Boston University, McGill University, and MIT.

Detection of transient ST segment episodes during ambulatory ECG monitoring.

Using the European Society of Cardiology ST-T Database, we have developed a Karhunen-Loève transform-based algorithm for robust automated detection of transient ST segment episodes during ambulatory ECG monitoring. We review current approaches and systems to detect transient ST segment changes and describe the architecture of our algorithm and its development. The algorithm incorporates a single-scan trajectory-recognition technique in feature space using the Mahalanobis distance function between the feature vectors. The main characteristics of the algorithm are detection of noisy beats, correction of the reference ST segment level to correct for slow ST level drift, detection of sudden significant shifts of ST deviation due to shifts of the mean electrical axis of the heart, detection of transient ST episodes, and, by tracking the QRS complex morphology, differentiation between ischemic and nonischemic ST episodes as a result of axis shifts. We compared the algorithm's performance to other recently developed algorithms and estimated its real-world performance.

Introduction of computational models to PhysioNet.

PhysioNet is a national research resource that provides experimental data sets and open-source software for their analysis. Computational modeling can complement studies of these experimental data sets so as to facilitate the advancement of physiologic research. Thus, in order to introduce computational models to PhysioNet, we have developed and posted a cardiovascular model designed for research that generates reasonable human pulsatile hemodynamic waveforms, cardiac output and venous return curves, and beat-to-beat variability. Some of the key features of the software include: 1) compatibility with PhysioNet's open-source data analysis software; 2) online viewing and parameter updating as the data are being calculated; 3) off-line viewing after completion of the simulation; 4) pre-compiled Linux binaries; 5) open-source code that may be compiled on other platforms; and 6) an extensive user's manual and software guide.

Karhunen-Loève transform as a tool to analyze the ST-segment. Comparison with QT interval.

The spatial and temporal courses of ventricular repolarization are quite sensitive to the biochemical and biophysiologic environment of the myocardial cells, and are therefore often an early marker of heart disease, particularly of ischemia. The detailed morphology of the surface electrocardiogram contains considerable information about the repolarization process. The ST-segment changes with ischemia, injury, and drugs. The QT interval is affected by drugs, heart rate, and autonomic tone, and in some situations may identify individuals at high risk for arrhythmias and sudden death. Variability in the shape, including duration, of the ST-T waves reflects autonomic nervous system activity and may identify high-risk patients. Automated methods for quantitatively characterizing ST-T complexes are important in studying long-term electrocardiographic records. Two computer-based measurement procedures for characterizing the repolarization period were comparatively analyzed: Karhunen-Loève (KL) transform representation of the ST-T shape and measurement of beat-to beat durations of repolarization (QT intervals). The results of KL transform representation and time-domain QT measurement algorithms for studying the repolarization period of the electrocardiogram on the European ST-T database are presented. It was found that about 20% of the records present a quasiperiodic KL pattern of ischemic ST-T activity and another 20% exhibit repetitive but not clearly periodic patterns of ischemic ST-T changes. From these ischemic records, 50% showed QT variations in at least one lead associated with the ischemic episodes.

Spectral characteristics of heart rate variability before and during postural tilt. Relations to aging and risk of syncope.

Fourier analysis of heart rate (HR) may be used to characterize overall HR variability as well as low- and high-frequency components attributable to sympathetic and vagal influences, respectively. We analyzed HR spectral characteristics of 12 healthy young (18-35 years) and 10 healthy old (71-94 years) subjects before and during 60 degrees head-up tilt. Total spectral power in the 0.01-0.40-Hz frequency range and low-frequency (0.06-0.10 Hz) and high-frequency (0.15-0.40 Hz) components of the HR power spectrum were significantly lower in old than in young subjects in supine and upright positions. To characterize and compare overall HR variability in young and old subjects, we computed the regression lines relating the log amplitude to the log frequency of the supine HR spectra (l/fx plots). The regression lines for old subjects were lower and steeper (mean slope, -0.78 [5%, 95% confidence limits (CL), -0.73, -0.83]) than in young (mean slope, -0.67 [CL, -0.62, -0.72]), indicating not only reduced overall spectral amplitude but also relatively greater attenuation of high-frequency HR components in the old subjects. This finding illustrates a novel way to quantify the loss of autonomic influences on HR regulation as a function of age. During postural tilt, HR variability was unchanged in the old subjects. For the entire group of young subjects, total HR variability increased during tilt. Six young subjects developed vasovagal syncope during tilt, enabling us to examine differences in the HR spectra of these subjects while they were asymptomatic before syncope.(ABSTRACT TRUNCATED AT 250 WORDS)

Cocaine alters heart rate dynamics in conscious ferrets.

This study was designed to test the hypothesis that cocaine intoxication induces distinctive alterations in sinus rhythm heart rate dynamics. Time-series and spectral analysis techniques were used to examine the effects of lethal doses of cocaine on heart rate variability in conscious, restrained ferrets. In all animals (n = 5), cocaine administration resulted in a marked decrease in sinus rhythm heart rate variability prior to sudden death. Heart rate variability (coefficient of variation of heart rate) just prior to death (0.018 +/- 0.005) was significantly (p less than 0.02) decreased compared to that at baseline prior to cocaine administration (0.061 +/- 0.022). There was also a significant (p less than 0.02) decrease in total spectral power prior to death compared to baseline. Transient low-frequency (0.04-0.10 Hz) oscillations in heart rate were also noted in three of the five animals following cocaine administration. There were, however, no significant changes in mean heart rate in response to cocaine. Alterations in heart rate dynamics were not seen in three saline-treated controls. Lethal effects of cocaine included ventricular arrhythmias (n = 2) and seizures (n = 3). One animal developed transient ST segment elevations that were consistent with coronary vasospasm. In conclusion, lethal doses of cocaine in the conscious ferret induce characteristic alterations in heart rate dynamics. These abnormalities (loss of heart rate variability and the appearance of low-frequency heart rate oscillations) are similar to those reported previously in certain patients at high risk of sudden cardiac death due to organic heart disease.

An automated, web-enabled and searchable database system for archiving electrogram and related data from implantable cardioverter defibrillators.

The advent of implantable cardioverter defibrillators (ICDs) has resulted in significant reductions in mortality in patients at high risk for sudden cardiac death. Extensive related basic research and clinical investigation continue. ICDs typically record intracardiac electrograms and inter-beat intervals along with device settings during episodes of device delivery of therapy. Researchers wishing to study these data further have until now been limited to viewing paper plots. In support of multi-center clinical studies of patients with ICDs, we have developed a web based searchable ICD data archiving system, which allows users to use a web browser to upload ICD data from diskettes to a server where the data are automatically processed and archived. Users can view and download the archived ICD data directly via the web. The entire system is built from open source software. At present more than 500 patient ICD data sets have been uploaded to and archived in the system. This project will be of value not only to those who wish to conduct research using ICD data, but also to clinicians who need to archive and review ICD data collected from their patients.

Do nonlinearities play a significant role in short term, beat-to-beat variability?

Numerous studies of short-term beat-to-beat variability in cardiovascular signals have not resolved the debate about the completeness of linear analysis techniques. This aim of this paper is to evaluate further the role of nonlinearities in short-term, beat-to-beat variability. We compared linear autoregressive moving average (ARMA) and nonlinear neural network (NN) models for predicting instantaneous heart rate (HR) and mean arterial blood pressure (BP) from past HR and BP. To evaluate these models, we used HR and BP time series from the MIMIC database. Experimental results indicate that NN-based nonlinearities do not play a significant role and suggest that ARMA linear analysis techniques provide adequate characterization of the system dynamics responsible for generating short-term, beat-to-beat variability.

Heart rate and respiratory rhythm dynamics on ascent to high altitude.

OBJECTIVE: To investigate the alterations in autonomic control of heart rate at high altitude and to test the hypothesis that hypoxaemic stress during exposure to high altitude induces non-linear, periodic heart rate oscillations, similar to those seen in heart failure and the sleep apnoea syndrome. SUBJECTS: 11 healthy subjects aged 24-64. MAIN OUTCOME MEASURES: 24 hour ambulatory electrocardiogram records obtained at baseline (1524 m) and at 4700 m. Simultaneous heart rate and respiratory dynamics during 2.5 hours of sleep by fast Fourier transform analysis of beat to beat heart rate and of an electrocardiographically derived respiration signal. RESULTS: All subjects had resting hypoxaemia at high altitude, with an average oxyhaemoglobin saturation of 81% (5%). There was no significant change in mean heart rate, but low frequency (0.01-0.05 Hz) spectral power was increased (P < 0.01) at high altitude. Time series analysis showed a complex range of non-linear sinus rhythm dynamics. Striking low frequency (0.04-0.06 Hz) heart rate oscillations were observed during sleep in eight subjects at high altitude. Analysis of the electrocardiographically derived respiration signal indicated that these heart rate oscillations correlated with low frequency respiratory oscillations. CONCLUSIONS: These data suggest (a) that increased low frequency power during high altitude exposure is not simply attributable to increased sympathetic modulation of heart rate, but relates to distinctive cardiopulmonary oscillations at approximately 0.05 Hz and (b) that the emergence of periodic heart rate oscillations at high altitude is consistent with an unstable cardiopulmonary control system that may develop on acute exposure to hypoxaemic stress.

The European ST-T database: standard for evaluating systems for the analysis of ST-T changes in ambulatory electrocardiography.

The project for the development of the European ST-T annotated Database originated from a 'Concerted Action' on Ambulatory Monitoring, set up by the European Community in 1985. The goal was to prototype an ECG database for assessing the quality of ambulatory ECG monitoring (AECG) systems. After the 'concerted action', the development of the full database was coordinated by the Institute of Clinical Physiology of the National Research Council (CNR) in Pisa and the Thoraxcenter of Erasmus University in Rotterdam. Thirteen research groups from eight countries provided AECG tapes and annotated beat by beat the selected 2-channel records, each 2 h in duration. ST segment (ST) and T-wave (T) changes were identified and their onset, offset and peak beats annotated in addition to QRSs, beat types, rhythm and signal quality changes. In 1989, the European Society of Cardiology sponsored the remainder of the project. Recently the 90 records were completed and stored on CD-ROM. The records include 372 ST and 423 T changes. In cooperation with the Biomedical Engineering Centre of MIT (developers of the MIT-BIH arrhythmia database), the annotation scheme was revised to be consistent with both MIT-BIH and American Heart Association formats.

Predicting survival in heart failure case and control subjects by use of fully automated methods for deriving nonlinear and conventional indices of heart rate dynamics.

BACKGROUND: Despite much recent interest in quantification of heart rate variability (HRV), the prognostic value of conventional measures of HRV and of newer indices based on nonlinear dynamics is not universally accepted. METHODS AND RESULTS: We have designed algorithms for analyzing ambulatory ECG recordings and measuring HRV without human intervention, using robust methods for obtaining time-domain measures (mean and SD of heart rate), frequency-domain measures (power in the bands of 0.001 to 0.01 Hz [VLF], 0.01 to 0.15 Hz [LF], and 0.15 to 0.5 Hz [HF] and total spectral power [TP] over all three of these bands), and measures based on nonlinear dynamics (approximate entropy [ApEn], a measure of complexity, and detrended fluctuation analysis [DFA], a measure of long-term correlations). The study population consisted of chronic congestive heart failure (CHF) case patients and sex- and age-matched control subjects in the Framingham Heart Study. After exclusion of technically inadequate studies and those with atrial fibrillation, we used these algorithms to study HRV in 2-hour ambulatory ECG recordings of 69 participants (mean age, 71.7+/-8.1 years). By use of separate Cox proportional-hazards models, the conventional measures SD (P<.01), LF (P<.01), VLF (P<.05), and TP (P<.01) and the nonlinear measure DFA (P<.05) were predictors of survival over a mean follow-up period of 1.9 years; other measures, including ApEn (P>.3), were not. In multivariable models, DFA was of borderline predictive significance (P=.06) after adjustment for the diagnosis of CHF and SD. CONCLUSIONS: These results demonstrate that HRV analysis of ambulatory ECG recordings based on fully automated methods can have prognostic value in a population-based study and that nonlinear HRV indices may contribute prognostic value to complement traditional HRV measures.