Electrocardiography in Ukrainian or know our

    Modern medicine is unthinkable without modern technology. And we unconditionally attribute these same technologies as a foreign country to America or Europe. It is not strange: ancient electrocardiographs from a neighboring hospital look like ridiculous clubfoot freaks next to modern, compact and functional imported devices.

    But in such an area as modern cardiography - equipment is only a flower.

    In fact, an electrocardiograph is a voltmeter with several channels and that’s it. But software ... software is an art.

    Imagine that a portable cardiograph was attached to a person - it is called a "Holter monitor" - in order to detect abnormalities in him, which occur every two days, and spontaneously (transient rhythm disturbances). A cardiogram is written all two days, saved. Everyone saw how much paper the cardiograph spends on a 10 second ECG in the nearest hospital? Now imagine that such an ECG was written 172800 seconds. And how do you order the doctors to look at so many changes, and not to miss something important? This is where modern software comes to the rescue.

    His responsibilities include analyzing the ECG and finding all the deviations on it.

    It becomes quite clear that now they rule not rings and magic, but tank wedges and carpet bombingsand high-precision analysis algorithms for rather complex and diverse ECG curves.

    - And why doctors? - you ask.

    And despite the fact that despite the development of technologies, hordes of programmers (in large companies up to 40 programmers are working on such projects), and considerable expenses of very large and respected companies, an absolutely accurate recognition algorithm has not yet been created.

    In general, the accuracy of the cardiogram recognition algorithm can be characterized by two parameters: sensitivity (Se) and specificity (Sp).

    The calculation of these parameters is based on statistics of mutually exclusive recognition events true positives (TP), true negatives (TN), false positives (FP) and false negatives Qo resent (FN).

    Se = TP / (TP + FN); Sp = TN / (TN + FP)

    Simply put: sensitivity is the ability to respond to the desired element, and specificity is the ability to prevent false positives.

    The accuracy of the algorithm for automatic analysis of cardiograms is determined during operational tests according to the ANSI / AAMI EC38: 2007 and EN 60601-2-47 standards (American and European, respectively).

    For this, a set of standard database of cardiograms is taken (AHA DB, MIT DB, ESC DB, NST DB, CU DB - in them each cardiogram is analyzed by a consilium of doctors and each element is described), cardiograms are converted into the format that this particular system “eats” to be checked, the system analyzes cardiograms in a fully automatic mode and creates an ECG description. An ideally accurate system should produce one hundred percent matching of the description for all cardiograms in all databases. There are no such algorithms at the moment.

    Official statistics, whose algorithm is more precise, do not exist, however, industry leaders (who have something to brag about) do not hesitate to publicize their successes. So familiarity with open sources (sites, apnotes, menus) allows you to make a comparative description of the best algorithms for automated ECG recognition:

    Analysis accuracy,%
    IMESC
    ECGpro
    Mortara
    veritas
    Agilent
    ST / AR
    Medilog
    ADAPT
    Spacelabs
    multiview
    MIT-BIH
    Q se99.9299.9399.6699.8799.82
    Q Sp99.8899.8599.8699.9199.85
    Aha
    Q se99.8499.8899.8099.9099.76
    Q Sp99.8099.8999.8799.9099.87

    This is where there is reason for pride, since one of them, namely ECGpro , from cover to cover from the smallest icon to the most complex algorithm was developed by the Ukrainian company IMESC ( www.imesc.com ), has been successfully living and developing more than 5 years. So much for modern medical technology.

    Know ours!

    And here are the screenshots:





    Also popular now: