Point Process Models of Human Heart Beat Interval DynamicsProvision and Dissemination of Software (Supported by NIH Grant R01HL084502) Back to Homepages: About this webpage This webpage will provide documentation about the state of the art of our point process algorithms for estimation of instantaneous indices of Heart Rate Variability (HRV). In addition to the C++ and Matlab Mfiles implementing the algorithms, we will post data sets which the user can download together with the programs. We will also maintain an active message board to respond to questions from users of the software, to post information about software updates and point to useful links related to cardiac monitoring. Making the software available via the web and creating an environment where users of the software can communicate will greatly facilitate software use and opensharing of experience with this analysis paradigm. Program framework In the final product the user will be able to specify input files with the RR interval series and the time series of eventual covariates, indicate a specific model (linear, nonlinear, covariates), the fitting algorithm (maximum local likelihood or adaptive filter), the desired time resolution for the output series, and the maximum orders of regression (or a fixed desired order). The output of the analysis will comprise: AIC and BIC, KolmogorovSmirnov plots and a summary goodnessoffit index suggesting the best order for the model, the time series of the instantaneous indices of HRV (mean RR and heart rate and respective standard deviations) computed for the order suggested by the best summary index, the autoregressive coefficients, and the eventual nonlinear and covariate indices computed at the same order. GUI development will allow for userfriendly selections and visualization of the results. A simplified beta version of the program can be downloaded below for demonstration purposes. Beta Version This program implements an adaptive AR(8) point process filter with a 5 ms time resolution applied to an RR interval series from a Tilt protocol (refer to Barbieri et al. 2006). Download Input Data file dd.txt (2 columns): time, RR interval (ms) Download MSDOS executable: AdaptiveFilter.exe. At this stage the program does not consider user interaction. Once launched, the program automatically creates three output files: 1) file HR_indices.txt (5 columns): time, mean RR, RR standard deviation, mean HR, HR standard deviation. Figure
1 shows the data display from file HR_indices.txt. Red segments are
referred to RAPID TILT, SLOW TILT, and STAND UP respectively (Barbieri et al, 2006).
2) file AR_coefficients.txt (10 columns): time, AR mean, AR coefficients.
3) file quantiles.txt (2 columns): time, empirical quantiles (t_{k}) . Figure 2. Empirical quantiles are ordered incrementally and plotted against the model quantiles to assess goodnessoffit (Barbieri et al, 2005).
Figure 3. The autocorrelation function of the empirical quantiles gives a further measure of goodnessoffit (Barbieri et al, 2005).
