Post on 05-Sep-2016
critical for assessing new treatments. Dynamic continuous ST-segment
recovery methods provide a quantitative marker of speed of reperfusion
with more information than serial static methods. However, continuous
data streams are hampered by data gaps caused by noise or artifact
despite optimal filtering. To overcome the impact of gaps on time until
first evidence of 50% or more reperfusion from peak ST levels (T1), we
used a novel statistical approach more attractive than simply eliminating
noisy data. If gaps occur before first observed recovery, T, then T may
or may not be T1. Our approach uniquely adjusts for variable data gaps
and multiple events.
Methods: A total of 610 continuous digital 12-lead ST-segment
recordings without gaps before T met criteria for analysis. Data gaps
were inputted into the gcleanh data streams with location and length
similar to historic ST-segment data. A gap likelihood function (GLF)
conditional on gap location was used to estimate T1. Standard statistical
methods are compared with the GLF using simulation methods to
compare efficiency, relative bias, and SD.
Results: See Table 1.
Conclusions: Gap likelihood function is more accurate than standard
statistical methods and provides more intuitive methodology for analyzing
continuous data streams.
doi:10.1016/j.jelectrocard.2005.06.058
Cosine-trend time-diversity electromagnetic interference filter for
electrocardiographic waveforms
Eric D. Helfenbein, A. Dean Forbes (Advanced Algorithm Research Center,
Philips Medical Systems, USA)
Digital electrocardiographic (ECG) recordings often contain unwanted
electromagnetic interference (EMI). Filtering techniques are usually
applied to remove the EMI so that the underlying physiological signal
can be processed. For high-resolution ECG applications such as
detection of afterpotentials or microvolt T-wave alternans, the major
challenge is to remove EMI while retaining the signal integrity. We have
designed a cosine-trend time-diversity EMI filter that has many
advantages over current EMI filtering methods and can be used for
both standard and high-resolution electrocardiography. The dominant
EMI in ECGs is usually the result of the sinusoidal power line and thus
usually appears as sinusoids at the power line frequency and a few of its
harmonics. A typical method for EMI removal is to process the
waveform with a digital notch or comb filter. Disadvantages of these
filters is that (1) they will bring,Q thereby introducing distortions; (2) they
remove signal of interest at the notch frequencies; and (3) they often do
not adapt to changes in power line frequency. The cosine-trend time-
diversity EMI filter takes advantage of gquieth regions between cardiac
cycles in the ECG signal during which time the EMI can be estimated.
First, the exact frequency of the EMI is determined using the fast
Fourier transform. Regression methods are then used to fit linear
combinations of basic functions composed of sine and cosine waves at
the power line frequency and its harmonics to the observed EMI. The
resulting estimate of the EMI signal is then subtracted from the ECG
waveform region of interest. We have applied this method to single
cardiac cycles using a sliding window approach and to longer regions
with excellent results. This novel filter method is highly adaptable to
variations in power line frequency and removes EMI with negligible
distortion to the underlying ECG.
doi:10.1016/j.jelectrocard.2005.06.059
Use of standards in the review of medical devices
Charles Ho, Donald Jensen, Frank Lacy, Neal Muni, Sabina Reilly,
Elias Mallis (Center for Devices and Radiological Health, Department
of Health and Human Services, US Food and Drug Administration,
Rockville, MD, USA)
The Center for Devices and Radiological Health of the US Food and Drug
Administration (FDA) uses a myriad of standards to facilitate the review of
premarket submissions of medical devices. The benefits of using standards
in this manner include providing a set of common requirements and test
protocols to the device manufacturer, thus reducing the manufacturer’s need
to breinvent the wheelQ each new bench test to ensure safety and
effectiveness of the device. Furthermore, with the present trend toward
international harmonization of standards, tests performed in accordance
with an international standard may be acceptable to several countries.
However, there are instances when the FDA does not agree with a few or
many provisions in a standard. This article aims to clarify the approaches
taken by FDA to balance or resolve disagreements. One approach begins
with the recognition of only some provisions of a standard or, more
commonly, excluding those parts of a standard that are unacceptable to the
FDA. Other approaches include working with the Standards Development
Organizations so that the standard can be revised to include a language that
is more agreeable to all parties involved. Specific examples will be
presented on medical devices such as electrocardiographic cables and
connectors, and noninvasive blood pressure monitors.
doi:10.1016/j.jelectrocard.2005.06.060
An objective test of T-wave alternans algorithm on noisy signals
Harry Hostetler a, Joel Xue b, Brian Young b, Willi Kaiser b, Martin Findeis b,
David Gutterman a ( aMedical College of Wisconsin, Milwaukee, Wisconsin,bGE Healthcare, Milwaukee, Wisconsin)
Background: T-wave alternans (TWA) has been associated with ventricular
arrhythmias. However, an objective test is needed to determine whether a
TWA signal can be identified in the presence of high levels of noise
generated during a typical exercise stress test when repolarization
variability may be augmented. A specific algorithm for isolating TWA
from electrocardiographic (ECG) signals during exercise was developed.
The purpose of this study was to test the hypothesis that the time domain
TWA algorithm can accurately quantify simulated exercise TWA patterns.
Methods: Noise was added to computer-generated ECG signals with
known TWA quantified by amplitude and phase. The TWA amplitude
varied from 0 to 100 lV measured as a peak-to-peak difference. The added
noise contained 3 basic types: high frequency with peak-to-peak voltage
varied from 20 to 500 lV, baseline wander varied from 0 to 2000 lV, andfiducial point shift from 0 to 16 m/s. This signal was then output to a
standard 12-lead ECG. Two systems were used to analyze this data: The
CASE system (GE Healthcare, Milwaukee, Wis), which is a standard stress
test machine, and a research workstation, which is a system that can run
experimental stress test algorithms. The experiment was repeated using
physiological noise from the MIT database. This noise was then applied to
the computer-generated ECG with known TWA. The TWA amplitude was
again varied from 0 to 100 lV.Results: From this, the CASE sensitivity in the time domain was found to be
75% at 9 lV, 83% at 38 lV, and 84% at 87 lV of alternans in combined
background noise. These peak-to-peak voltages are equivalent to 4, 10, and
20 lV, respectively, of mean TWA as measured in the frequency domain.
Table 1
Method Parameter Relative bias SD Median
True values 0.0168 – – 41
Exclude data with gaps 0.0232 0.34 0.005 30
Standard interval censor 0.0189 0.12 0.002 37
Standard right censor 0.0176 0.05 0.002 39
GLF 0.0171 0.00 0.002 41
Poster Session II / Journal of Electrocardiology 38 (2005) 88–93 89