Rhythm detection system in AED mode
102 / 116
DefiMonitor XD 21324 / GB / K05
METRAX GmbH
specific amount of time, the device will rate the heart rhythm shockable. However,
if one of the parameters exceeds its higher threshold value, this shocking
recommendation is retracted.
•
Filters and measures artefacts and interference. If a specific artefact level is
exceeded no reliable signal analysis is possible and the heart rhythm is rated non-
shockable.
•
Detects pacemaker pulses and removes them from the ECG signal before
analysing rhythm.
•
Measures the heart rate. If it drops below a threshold, the heart rhythm is rated
non-shockable.
Cardiac rhythms used to test the rhythm detection system in the device
To validate the databases used:
•
AHA Database for Evaluation of Ventricular Arrhythmia Detectors (80 records
with 35 min length each)
•
MIT-BIH Arrhythmia Database (48 records with 30 min length each)
•
European ST-T Database (48 records with 120 min length each)
These databases contain ECG rhythms with ventricular fibrillation (VF) of varying amplitudes,
ventricular tachycardia and sinus rhythms with
–
amongst others
–
supraventricular
tachycardias, atrial fibrillation/flutter, sinus rhythm with premature ventricular contractions
(PVC), asystolia and pacemaker pulses. All records are valid for evaluation of the VT-/VF-
rhythm detector with respect to used electrode systems and ECG signal processing
characteristics. For validation the ECG was converted to an analogue signal and put out to the
device under test. The detection result was than read back by the validation system and
compared to the reference annotation.
As shockable rhythms, in calculation, we look at the characteristic values of the sections in the
ECG datasets of the above databases, which are marked with the PysioBank Annotationscode
for ventricular flutter ("[" Begin, "]" End; also refer
) using the PhysioBank
Annotationscode.
These sections also contain ventricular tachycardia that however are not separately annotated
and therefore cannot form part of the statistic.
For evaluation, the test and reference annotations of segments with a length of 12 seconds
were compared to each other. Segments containing a shift between shockable and non-
shockable rhythms were not taken into account.
Performance results (weighted average):
Sensitivity
99,30 % (Requirement of IEC 60601-2-4:2003: >90 %)
Specificity
99,88 % (Requirement of IEC 60601-2-4:2003: >95 %)
False positive rate
0.04 %
Real predictive value
97,93 %
The databases used have a total length of about 10,000 minutes. The calculation was made in
accordance with IEC60601-2-4-2003.