Poster
# 61
Poster |
|
6th Internet World Congress for Biomedical Sciences |
Marcelino Martinez_Sober(1), Juan Guerrero(2), Gustavo Camps i Valls(3), Antonio J. Serrano López(4), José David Martín Guerrero(5)
(1)Universidad de Valencia - Burjassot. Spain
(2)Departamento de Electrónica. Universidad de Valencia - Burjassot. Spain
(3)Universitat de València - Burjassot, Valencia. Spain
(4)Dpto. Electrónica. Universidad de Valencia - Burjasot. Spain
(5)G.P.D.S. Departament d´Enginyeria Electrònica. Universitat de València - Burjassot. Spain
|
 |
|
|
[Cardiolovascular Diseases] |
[Health Informatics] |
ECG is the graphic representation of the heart electrical activity. During the heartbeat, a well-known pattern of electrical changes reflects exactly the action of the heart. This activity can be picked up through electrodes placed on the surface of the body. The activity of the heart is represented by characteristic waves and can be evaluated in any media support: screen visualisation, printed copy, or stored for its later study.
On the other hand, Holter tapes, through the ECG of a patient can be registered during the 24 hours of the day, has become a very useful tool for detection and pathologies modelling. The patient holds the electrodes attached to the body and is connected to a box that includes a recording device. At that moment the information is stored in an analog way into an audio tape that should be sampled for a later treatment.
The digitalisation of Holter tapes is carried out by means of a continuous system of acquisition constituted by a system of reading the tapes (Pathfinder), a visualisation application and a data acquisition card PCL711B (fig. 1). By means of our visualisation program the relative acquisition parameters are controlled (gain, channel, sampling rate etc.).
The acquisition time of each tape is 23 minutes (corresponding to 23 hours, because recording speed is 50 times slower than acquisition Holter speed) so the size of each one of them is around 40 MB. The storage of the files in digital support is carried out on CD-ROM units because they are really extensive files.
As regard the series, two periods have been analysed:
-Diurnal period: with a duration of 6 hours starting from 9:00am o´clock
-Nocturnal period: with a duration of 4 hours starting from 1:00am o´clock
FIDUCIAL POINT MARKING PROGRAM
For the marked of pulses and visualisation of the digitised signs was carried out an application in visual environment taking advantage of the versatility of the programming GUI of Matlab 5.0 (2).
I. Process of Marking Pulses.
As we say former, in the studies of time series of 24 hours, 2 periods are distinguished one day and another nocturne that they will be analysed for separate. In the process of having marked a set of waves, an ECG is shown so that the cardiologist locates a tract of the sign of good quality or that in which the patient´s pathology is observed with more clarity. Once selected the period, the program visualises in separated mode each of the pulses and invites the cardiologist to mark the points of interest just as it is shown in fig. 2
The marking process is repeated until marking the desired number of points. When concluding this process of marked for the day period and nocturne is generated a file of marks that will be used later on.
Once located the areas of those that we want to obtain the series RR, we proceed to use a detection program of this series implemented in C language. For the QRS complex localisation we have used the detector included in the MIT-BIH arrhythmia database, developed by W.A.H Engelse and C. Zeelemberg (1). The algorithm uses a band-pass filter to attenuate the frequencies outside of the band of the QRS complex. To be able to locate the areas of ECG that are possible candidates to QRS, thresholds they are used. These are calculated starting from the file of having obtained in the previous stage and parameters of the acquisition like the resolution of the converter, gain etc.
After the visualisation of some of the obtained series, it has been checked that some recorded files can contain errors due to losses of contact of the electrodes or a bad quality of the signal. In such cases detector produces a high number of false detections.
Under these conditions, whenever the quality of the signal in adjacent tracts allows them, it is preferable to redefine the day and nocturne so-called periods with the purpose of having more appropriate temporary series. Really this doesn´t affect since to the later studies with these periods we are distinguishing among activity/quietness for what the time limit considered is not decisive.
To be able to carry out that indicated previously it is much more convenient not to consider any period type (day/night) and to obtain the temporary series from the registration to the complete, such and like it is shown in fig. 3.
If in the RR time-series picks outside of what we can consider as the mean value of the whole RR series is observed, it indicates us that the detector of QRS is not working properly. This can be due to that the signal is of very bad quality, or simply to that the detector has its limitations. In these cases the signal visualisation with the marks facilitates us the determination of what it is happening. As the visual inspection of all the periods is very tedious, exists a program that analyses the series and determines if the number of false detections is acceptable or not.
The time series are stored in order to be analysed and to obtain parameters of interest in the frequency and time domain that will serve as indicators in the determination of heart pathologies.
The study of time series in Cardiology is applied defining the occurrence of the wave R of the electrocardiogram (ECG) as an event, and the statistics associated to the interval among pulses (interval RR) are studied. We talk then of variability of the heart rhythm (HRV). The study of the HRV is usually carried out with short registrations (from 2 to 5 minutes) and long ones (24 hours: Holter), and it provides information on the operation of the physiologic mechanisms of regulation of the HRV. In the spectrum of the HRV they can be distinguished 4 bands:
1.-ULF: associated with the modulation of the rhythm to hormonal influence, etc. Cycles from 2.8 hours to 5.6 minutes).
2.-VLF: associated to the temperature, etc. Cycles from 5.6 minutes to 25 seg.
3.-LF: associated to the vegetative system. Cycles from 25 to 6.7 seg.
4.-HF: associated to the breathing cycles from 6.7 seg to 2.5 seg.
In this work a program developed for the standard study (3) of series RR Holter is presented.
The program for processing the signal has been developed with Matlab v4.2. The one processed is carried out in the following phases:
1.-Detection of non valid tracts and anomalous pulses (echtopics): The echtopic detection is made checking a value of more joining that 20% of the previous RR, or a value outside of the range RR_min - RR_max. If the number of data to eliminate is superior to a maximum value (usually, 20%), the series is rejected.
2.-Elimination of non valid periods and echtopics: It is eliminated the echtopic pulse and the following one (compensatory pause), and it is substituted by values RR obtained by some of the methods: to) Eliminate and Paste; b) Lineal Interpolation; c) Splines Interpolation; d) AR predicction. The relative number and the relative duration of the intervals omitted or interpolated RR are saved (4). Figure 4 fig. 4 shows an original RR series with echtopics and non-valid tracts, and after eliminating both types.
3.-Division in short series: the series is divided in tracts of 5 ´. The study in the frequency domain is carried out averaging the values of the bands LF and HF calculated for each subseries of 5 ´, and the obtained sampled series is undersampled to calculate VLF.
4.-Sampling of the five minutes series. Due to the RR series is not uniformly sampled a continuous signal in the time that is later on regularly sampled before carrying out the spectrum study is obtained by means of interpolation. The program admits polynomials interpolation methods of order 0, 1 or splines.
5.-Elimination of tendencies and calculation of the spectrum with Welch method.
6.-Calculation of parameters in the domain of the frequency for series of 5 ´: to) MLF (Frequency of the maximum in LF); b) MHF (Frequency of the maximum in HF); c) PLF (Power in the band LF); d) PHF (Power in the band HF); and) PLFN (normalised Power PLF/(PLF+PHF)); f) PHFN (normalised Power PHF/(PLF+PHF)); g) LF_HF (PLF/PHF rate).
7.- Undersampling of the series of 5 ´ to study the band VLF of the total series.
8.-Obtaining of the total spectrum and parameters in VLF: h) MVLF (Frequency of the maximum in VLF); i) PVLF (Power in the band VLF).
9.-Calculation of the parameters in the time domain: ANN (mean RR), SDNN (std of all the RR), RMSSD (value RMS of the differences among adjacent RR), CV (variance coefficient: SDNN/ANN), SDANN (std of the ANN of the subseries of 5 ´), Triangular index (integral of the histogram / maximum value).
A series of programs in Matlab that allow us to automate the process of obtaining of RR time series RR of heart registrations have been developed. Excepting the control of the digitalization of the Holter tapes, which is made by means of a program in C under MS-DOS, and the function that determines the RR series, the rest of functions has been developed in Matlab.
A program has been developed for the study of RR time series. The program allows obtaining the parameters standard in the time and the frequency domain for series obtained from Holter recordings.
To conclude we will only indicate that the following step would be the inclusion of the necessary modifications in the algorithm so that besides the RR series is possible to obtain another type of series of great clinical interest, as they are the QT series.
This work has been partially sponsored by the project GV97-TI-05-14 of the Generalitat Valenciana in the mark of the Program of Projects of Scientific Investigation and Technological Development."
|
|
|
|
|
|
[Cardiolovascular Diseases] |
[Health Informatics] |
