STAR arrhythmia monitoring.pdf

7/29/2019 STAR arrhythmia monitoring.pdf

1/16

ST/AR Algorithm Software

ST/AR ArrhythmiaMonitoring

Application Note

The ST/AR (ST and Arrhythmia) algori thm is a multi -

lead ECG algori thm designed for both arrhythmia and

ST segment monitori ng.The arrhythmia algorithm can

run as single-lead analysis or multi -lead analysis. Thisapplication note provides a detailed descripti on of the

arrhythmia monitoring port ion of the ST/AR

algorithm. T his document also descri bes the arrhythmi a

systems alarm structure. The assessment of the

arrhythmia algorithms performance is described in a

separate publicati on ti tled Assessing ST/AR Arrhythmi a

Performance wi th publication number 5980-1199E.

IntroductionComputerized arrhythmia monitoring is a valuableclinical tool in many patient areas. To be most

effective, this tool requires a thorough knowledge ofthe systems features, how the computer processesthe signals, as well as the proper applicationprocedures.

The ST/AR arrhythmia monitoring algorithm isdesigned to process one or two simultaneouschannels of surface ECG signals for detectingchanges in the ECG rhythm while offeringcontinuous patient surveillance and alarmgeneration.

Through a sophisticated computer algorithm, QRScomplexes are detected, labeled and classified. Basedon the classification, the computer then generatesalarms. It is the intent of this application note toexplain the fundamentals involved in each of thesesteps.

Arrhythmia MonitoringAlgorithmAn algorithm is a set of rules and instructions thatcomputers use to analyze data. The arrhythmiamonitoring algorithm processes the ECG signals for

both paced and non-paced patients.The algorithm performs several actions on theincoming ECG waveform, including filtering thesignal, detecting and classifying the QRS, generatingheart rate, identifying ectopic events and rhythms,and generating alarms if necessary.

N N VV N


2/16

2

Quality Check of the ECG Signal

Before monitoring begins, the ECG signal quality ischecked for noise and inoperative (INOP)conditions.

Noisy ECG Signals

Noise refers to any degradation of the ECG signal

that makes it difficult to accurately detect andclassify beats. Causes of noise, such as artifact andelectrical interference, should be avoided wheneverpossible.

The following are some possible causes of noisyECG signals:

Poor skin prep.

Dried electrode gel.

Detached electrodes.

Broken lead wires.

Muscle artifact caused by shivering, movement ortremors.

Baseline wander caused by excessive chestmovement, or the offset differences between twobrands ofelectrodes.

Respiration artifact caused by thoracic orabdominal movement of both spontaneous andventilated breathing patterns.

Equipment.

Prompt attention by the clinician to any of the aboveECG interferences increases the accuracy of thealgorithm and decreases the incidence of falsealarms. The causes of noisy signals and possiblecorrective actions are shown in the table on thefollowing page.

INOP Conditions

Inoperative conditions which interfere with orprevent monitoring the ECG signal can alsointerfere with arrhythmia monitoring. A leads offcondition which results in the loss of ECGmonitoring will also inhibit arrhythmia monitoringuntil the condition is corrected and a lead has beenrestored. If using a 5-lead wire set, whether doingstandard or EASI lead placement, a leads offcondition does not necessarily result in the loss ofmonitoring. The arrhythmia algorithm will usewhichever lead(s) are available for monitoring.

Multi-Lead Monitoring

While in most cases highly accurate results areobtained when monitoring two leads of ECGsimultaneously, it is important to remember thatboth leads of ECG are being used for detection,classification and alarm generation. The quali ty ofboth signals will effect the accuracy of thearrhythmia algorithm in beat detection,

classification, and alarm generation.Even though a multi-lead arrhythmia algorithm hasbetter ability in handling noisy signals than a single-lead algorithm, in order to achieve the maximumperformance it is important that the two ECG leadsselected for monitoring be free of noise.

In the following example, the second lead isextremely noisy, and therefore it provides little valueto QRS detection. During classification both leadsare used. The second noisy lead may impactnegatively on the final beat classification. In

addition, i f the channel 1 ECG becomes inoperative,the second lead will be the only lead available foranalysis; hence poor performance will result.

Although the ST/AR algorithm has an improvedhandling of noisy signals and the changingamplitudes caused by the loss of a good lead, it is stillimportant to choose the best two leads available. Ifthere are false alarms, examine both leads. You mayneed to select a different lead or change theelectrodes or electrode position if there is excessivenoise, unstable voltage, low amplitude or large P- or

T-waves.

In cases where selecting a different lead or changingelectrode position to correct the problem is notpossible or practical, then it is better to select thelead with best signal quality.


3/16

3

Noisy ECG Problem Solving

Problem Appearance Cause Treatment

Power Line Interference (50 /60-Cycle Interference)

Regular sawtoothbaseline withexactly 10 peaksevery 5 mm at 25mm/sec. (50 cycle)

or 12 peaks every 5mm at 25 mm/sec.(60 cycle).

Poor electrode placement.

Possible non-groundedinstrument near patient.

Reapply electrodes.

Disconnect electrical appliancesnear patient (one at a time) by

pulling wall plugs, to determinefaulty grounding. Have engineercheck grounding.

Muscle Artifact Fuzzy, irregularbaseline.

Tense, uncomfortablepatient.

Poor electrode placement.

Tremors.

Diaphoresis.

Make sure patient is comfortable.

Check that electrodes are appliedon flat, non-muscular areas of thetorso; reapply electrodes ifnecessary.

Irregular Baseline Rough, jaggedbaseline.

Poor electrical contact.

Respiratory interference.

Faulty electrodes.

Dry electrodes.

Reapply electrodes, using propertechnique.

Move electrodes away from areaswith greatest movement duringrespiration.

Apply new electrodes.

Baseline Wander Rhythmic up-and-down movement of

the ECG baseline.

Movement of the patient.

Improperly appliedelectrodes.

Respiratory interference.

Make sure the patient iscomfortable.

Reapply electrodes. Check thatpatient cable is not pulling onelectrodes.

Move electrodes away from areaswith greatest movement duringrespiration.

Poor Electrode Contact Trace switchingfrom high to low in

steps.

Loose electrodes.

Defective cables.

Change all electrodes, using goodskin prep.

Replace cables.


4/16

4

ECG Analysis

Step 1: ECG Signal Filtering

Digital Sampling

The patients incoming ECG waveforms are digitallysampled at 500 samples/second. This high ECGsampling rate is used to preserve the narrow pace

pulses (spikes) for accurate pacing analysis. For QRSdetection and ventricular fibrillation detectionwhere high sampling rates are not needed, a lowersampling rate of 125 samples/second is used. ForQRS classification, a sampling rate of 250 samples/second is used. The sample rate for the M300A andM3001A Measurement Server is 125 samples/second for QRS classification.

Pace Pulse Processing

With both bedside monitors and telemetry systems,

detection of the pace pulse for patients withpacemakers occurs in the bedside monitor or thetelemetry transmitter. This permits highly accuratepace pulse detection on the unfiltered ECG signal.Using the 500 samples/second data, the pace pulsesare eliminated before the ECG waveforms areprocessed by the QRS detector. In this way theaccidental detection of the pace pulse as a QRS isprevented.

Note: The removal of pace pulses is done only onthe signals that are used by the algorithm forarrhythmia analysis. The pace pulses remainavailable for viewing on the display or recordings.

Filtering

Next, the ECG waveform(s) are processed by twodigital filters: a detection filter and a classificationfi lter. These filters are optimized individually toenhance the performance of QRS detection andclassification.

a. Detection Filter

The detection filter removes low frequency noise(baseline wander) and muscle artifact, andaccentuates the QRS complexes. P-waves and T-waves are diminished. This filter makes it easier toaccurately detect the QRS and helps avoiderroneously detecting tall T-waves or artifact asbeats. Since it distorts the true shape of the QRS, theoutput from the detection filter is used only for beatdetection.

A special filter is used for neonatal ECG processing.This filter improves detection sensitivity of narrowneonatal QRS complexes.

b. Classification Filter

The classification filter also removes signalirregularities, but it preserves the important featuresof the QRS. Since this filter does not distort thecomplex, the resulting ECG output can be used forfeature measurements and beat classification.

Step 2: QRS Detection

The algorithms challenge in QRS detection is tofirst locate R-wave peaks that becomecandidatepeaks and then to make sure that they are notactually noise, P- or T-wave peaks.

ECG Amplitude

In order to comply with AAMI EC-13 specification,ST/AR internally removes the gain adjustmentsbefore the signal is analyzed for detection andclassification. The detection threshold for the QRS

cannot be less than 0.15 millivolts. Thisspecification is aimed at preventing the detection ofP-waves or baseline noise as QRS complexes duringcomplete heart block or asystole.

Thus increasing or decreasing gain at the point-of-care device has no effect on the ECG size used forQRS detection. The algorithm will analyze the ECGsignal as it would appear at a gain x 1. Therefore, foroptimal performance and to prevent false alarmssuch as pause or asystole, it is important that thelead(s) selected for monitoring have adequateamplitude. This can be confirmed by comparing the

ECG signal to the one-millivolt reference bar on thedisplay and recordings.

The bedside monitor and Information Center

displays, as well as recordings, will show a waveformwith the gain adjustments and a one-millivoltreference bar at the beginning of the waveform.Consequently the clinician can be looking at awaveform which appears large until compared to theone-millivolt reference. ST/AR sees a smallerwaveform (gain x1) which it analyzes.

One-millivolt bar withGain x 4:

What the clinician sees on the display - gain appliedto signal

One millivolt bar withGain x 1

What the ST/AR algorithmsees after removingthe gain.


5/16

5

Combining Multiple Leads Into A SingleSignal

With multi-lead analysis, after both ECG signalspass through the detection filter, they are combinedinto a single signal for QRS detection. Thecontribution from each ECG lead to the QRSdetection signal is proportional to its measuredquali ty based on the waveform amplitude, and the

amount of muscle and baseline noise. The weightingfactors are updated at least every 200 milliseconds toallow for quick adaptation to signal quality changes.

The QRS detection signal can dynamically adapt tothe quality of the incoming ECG signal(s), thus theimpact of noisy signals to QRS detection can bereduced.

Generating the QRS Detection Signal Using TwoECG Signals

Locating Candidate R-Wave Peaks

The QRS detector checks the QRS detection signalfor the presence of the peak of an R-wave. Searchingbegins after an absolute refractory period from thepreviously identified QRS complex. This helpsprevent a T-wave from being identified as an R-wave. The value used for the absolute refractoryperiod is 192 milliseconds for adult and pediatricpatients. A smaller value, 160 milliseconds, is usedfor neonatal patients.

A moving search region is established at the end ofthe refractory period. For each search region, a newthreshold is established based on:

Noise around the search region.

Distance from the previously detected R-wave.

Averaged R-wave height.

The largest peak within the search region isconsidered a candidate R-wave.

Minimum Detection Threshold

To prevent the detection of P-waves or noise as QRS

complexes during complete heart block or asystole,the detection threshold will never go below thelarger of 1/5 of the average R-wave height or 0.15

millivolts. Any peak smaller than this value is notdetected.

Noise Rejection

After a candidate peak is detected, it is checkedagainst a noise threshold to make sure that it is not anoise artifact or a QRS complex surrounded bynoise. If the ECG lead(s) are determined to be noisy

then a beat label A is assigned to the candidatepeak and no classification is performed. With multi-lead, the noise check is performed on each leadindependently. Only the lead that is identified asnoisy will be excluded from subsequent analysis.

Peak Rejection

Before accepting the peak as a potential R-wave,there are two further tests which are carried out oneach lead independently.

a. Potential False Identification of the P-Wave

To prevent a P-wave that is associated with a QRSfrom being counted as a QRS, it is checked againstwhat is known about the previously identified P-waves. If it is found to be similar, the peak isrejected.

b. Potential False Identification of the T-Wave

If a candidate peak is found in close proximity to thepreceding beat, it is tested to see if it might be a lateT-wave. After a series of height and timing tests, thepeak may be determined to be a T-wave and

rejected.

If it is determined that the candidate peak is neithera P-wave or a T-wave, it is identified as a QRScomplex and saved.

P-Wave Detection

After a QRS complex is located, a search is made oneach lead independently in the area prior to the beatto determine if there is an associated P-wave. Thisarea is 200 milliseconds wide (104 milliseconds forneonate) and ends 120 milliseconds (56 milliseconds

for neonate) before the R-wave peak. To beconsidered a P-wave, it must be at least 1/32 of theR-wave height and the P-R interval must be close tothe average P-R interval.

In other words, the candidate P-wave must representaverage characteristics in its relationship to the QRS.P-wave detection is used to differentiate between aSinus Rhythm (normal QRS complexes withassociated P-waves) and a Supraventricular (SV)Rhythm (normal QRS complexes without associatedP-waves).


6/16

6

P-Wave Detection for Adult/Pedi

Step 3: Feature Measurement

After a beat is detected, it is measured in a numberof ways to determine its features. These featuresrepresent beat characteristics which can be used todiscriminate between different types of beats.

The features measured are: height, width, area, andtiming (a series of R-R interval measurements).With multi-lead analysis, the height, width and area

are measured for each lead independently.

Step 4: Beat Labeling

Once the signal quality is checked and verified, andthe QRS is detected and measured, the beat islabeled. Labeling means the algorithm assigns thecomplex one of the following labels:

If the signal quali ty is not good, the algorithmassigns one of the following labels to the waveform:

Another label used to label the ECG signal is:

Beat labeling involves three major concepts:

1.The use of template families to represent recurringmorphologies.

2.The use of initial learning of the patient's normal

morphology and/or paced morphology if thepatient is paced.

3.The use of template families to aid classification ofQRS complexes.

Grouping into Template Families

To aid the algorithm in labeling a new beat,previously detected beats that have similar shapes aregrouped into template families.

Each template family contains the followinginformation:

Template family classification: Normal,Ventricular, Paced, or Unclassified.

Waveform template, generated by averaging all thebeats that are considered similar enough to beincluded.

The number of complexes having this shape.

The length of time since this shape was last seen.

Statistical information on the beats feature

measurements.

If the patient is paced, pace pulse informationassociated with the beats.

For each patient, up to 16 different active templatefamilies can be created for each individual lead (withthe M3/M4 monitor, it is up to 8 templates). Tokeep the template family information current, theyare dynamically created and replaced as the patientsbeat shapes change.

If the patient begins to display a new beatmorphology, a new template family is created. Oldertemplate families from beats the patient is no longerexperiencing are automatically deleted.

Example of Template Families

N Normal

S Supraventricular

PrematureV Ventricular Ectopic

P Paced

? Questionable

L Learning

I Inoperative

A Artifact

M No QRS

Template FamiliesExisting

ECG To Be Analyzed


7/16

7

Template Matching

When a beat is detected, it is matched against thestored waveform templates for that patient.Matching means comparing the beat shape with awaveform template. This process involves overlayingthe beat on the template and using a mathematicalprocedure to measure the differences between thetwo shapes.

Learning

When arrhythmia monitoring starts, a learningprocess is initiated. The goal is to learn the patient'snormal complexes and/or the paced complexes if thepatient with a pacemaker is in paced rhythm. Thelearning process involves the first 15 valid (non-noisy) beats encountered during the learning phase.

The family selected to represent thenormalincludes the beat that is the most frequently seen,narrowest, on-time beat. For this reason, learning

should not be ini tiated when the patient's rhythm isprimarily ventricular.

A manual relearning should be initiated if beatdetection is not occurring or if beat classification isincorrect, and results in a false alarm. Remember,however if the same signal condition exists whichcaused the algorithm to perform poorly, relearningwil l not be able to correct the problem. The problemcan only be corrected by selecting a different lead.

Learning for Single-lead Monitoring

When single-lead analysis is selected, the algorithmlearns the primary ECG lead whenever:

ECG monitoring is init iated with arrhythmia

turned on.

Arrhythmia monitoring is turned on.

The relearn key is activated.

Pacer status is changed.

After aLeads Off INOP situation lasting longerthan 60 seconds has reversed.

The ECG Lead or Lead Label is changed.

Learning for Multi-lead Monitoring

When multi-lead analysis is selected, the algorithmlearns both ECG leads simultaneously whenever:

ECG monitoring is init iated with arrhythmiaturned on

Arrhythmia monitoring is turned on.

The relearn is activated.

Pacer status is changed.

For minimum interruption of continuousmonitoring the ST/AR algorithm provides theflexibility for learning each lead independent ofanother lead. The algorithm continues monitoringone ECG lead while the other lead is being learned.The unaffected lead will be analyzed continuouslywithout interruption.

The algorithm learns only the affected leadwhenever:

The ECG lead is changed.

A new ECG lead becomes available.

A prolonged Leads Off situation is reversed.

Beat Labeling

Once the initial learning process is over and anormal template family is established, each newlydetected beat is labeled:

1.If the beat matches a waveform template of a familywhich has already been classified:

- The beat is labeled the same as the template

family.

- The template is updated, the population isincreased, and the new beat features are added tothe familys statistics.

2.If the beat matches a waveform template of afamily which has not been classified:

- The beat is labeled. using the beat labeling rules(described below).

- The template family is updated, and whenenough beats matching the template occur, the

template family is classified using the templatefamily classification rules (described below)

New Beat

Stored Template 1 Stored Template 2

Poor Match Good Match


8/16

8

3.If the beat does not match any of the existingwaveform templates:

- The beat is labeled using the beat labeling rules.

- A new template family is created with this beatas its first member.

Beat Labeling Rules

Every beat is analyzed using thebeat labeling rules.These rules determine a preliminary label for eachbeat:

To help avoid artifact being labeled V, a new beatshape may be labeled as ? when initially seen.

The beat labeling rules use a combination of thefollowing information:

Feature measurements.

Timing/Rhythm.

Template matching.

Morphology similarity to neighboring beats.

Pace pulses associated with the beat (if paced

patient).

The rules emulate the behavior a clinician uses whenanalyzing an ECG waveform. Even though the samerules are used for both the adult and neonatemonitoring modes, several thresholds are adjustedfor monitoring neonatal patients to account for theirhigher heart rate and narrower QRS complexes.

Pace Beat Classification

First, the algorithm searches for both atrial andventricular pace pulses. To accomplish this, asearch

window is established prior to the QRS complex.Then, as pace pulses are seen in the search window,their distances from the beat are tracked.

For the algorithm to consider the new beat paced,the pace pulses to QRS distance must be similar tothe pace pulses associated with the paced template.

If the pace pulses fall at random distances they areconsidered to have no effect on the beat. Thedistance that the pulses are found from the beatdetermines the type of pacing - atrial, ventricular orAV (atrial/ventricular) paced.

.

Template Family Classification Rules

After the normal family is initially learned, eachnewly created family remains unclassified untilenough similar beats, usually less than five, are seento classify it accurately.

Template families are classified as:

The labels given to the beats using the beat labelingrules are used to determine the family classification.

A normal template can include N and S beats. For atemplate to be classified as V, a majority of membersmust be labeled V. By the ninth beat, if the templateis not classified as V, the template is classified N.

A paced template is determined by statistical analysisof all the pace pulses detected within 600milliseconds of the QRS complexes that are includedin the template family. In order for the algorithm toconsider the template paced, pace pulses must fall at

N Normal

S Supraventricular Premature

V Ventricular Ectopic

P Paced

? Questionable

N Normal

V Ventricular EctopicP Paced

Atrial

Pacing

Ventricular

Pacing

Dead

Zone

Random

Paced Pulse

Paced

Template

600 msec 150 msec 48 msec

(30 msec for neonate)

Random Pace Pulses Around QRS

Clustered Pace Pulses Around QRSClustered PacedPaced Pulse Template

Pace Pulse Search Window


9/16

9

a consistent distance from the QRS. Based on thenumber of consistent distances found and theirvalues, the paced template is classified as atriallypaced, ventricular paced or AV paced.

The template family classifications are continuouslychecked against labels given to the individual beatsusing the beat labeling rules. If many discrepanciesare found, the template family is reclassified. In this

way, the algorithm has a mechanism to correctinaccuracies made during template classification.

Beat labeling rules use current information toanalyze ectopic activity, while template matchingprovides long term memory to the algorithm. Byusing this combination of beat features, timing, andtemplate matching techniques, the algorithmflexibly manages a variety of conditions with a highdegree of accuracy.

Step 5: Ventricular Fibrillation

DetectionWorking in parallel with beat detection and labeling,a separate detector continuously examines the in-coming ECG signal(s) for ventricular fibrillation. Ifa flutter or sinusoidal wave pattern persists for morethan four seconds in both ECG channels, themonitor alarms for ventricular fibrillation.

If single-lead arrhythmia monitoring has beenselected, only the primary ECG is used for thedetection of ventricular fibrillation.

Step 6: Rhythm and AlarmDetection

The results from beat labeling and ventricularfibrillation detection are used by the rhythm andalarm detector as it measures the heart rate,determines the patients underlying rhythm, andidentifies ectopic events.

Alarms are activated by the alarm generator. Higherpriority alarms, such as asystole, take precedence andsupersede lower priority alarms, such as low heart

rate.

Heart Rate Computation

Two different averages are used by the arrhythmiaalgorithm to determine the heart rate:

Normally, the heart rate is computed by averagingthe most recent 12 R-R intervals. Beats N, S, P, andV are all included in the computation. This averagegives a stable estimate of the underlying heart rateeven when the rhythm is irregular.

When the heart rate drops below 50 bpm (80 bpmfor neonates), the number of R-R intervals used inthe average is dropped to four to improve the

response time for the computed heart rate to reachthe correct value during bradycardia.

Note: For the ventricular tachycardia alarms whichhave a user-definable PVC run length limit, theheart rate is computed based on the user selectedPVC run length up to nine PVCs maximum (i.e. upto 8 R-R intervals). For instance, if the VT alarm isuser-defined as five or more consecutive PVCs and

heart rate greater than 100 bpm, then four R-to-Rintervals will be used to compute the heart rate to seeif the rate exceeds the limit of 100 bpm. Likewise, ifthe VT alarm is user-defined as ten or moreconsecutive PVCs and heart rate greater than 100bpm, then eight R-to-R intervals will be used as thatis the maximum possible.

Arrhythmia Monitoring for the PacedPatient

Detection of the pace pulse occurs at the point of

care device (bedside monitor or telemetrytransmitter). This permits highly accurate pace pulsedetection on the unfiltered ECG signal.

Since the pace pulses are detected and eliminatedBEFORE the ECG waveform is processed by theQRS beat detector, accidental detection of the pacepulse as a QRS is prevented. The pace pulses remainavailable for viewing on the display or recordings.The pace pulses are displayed at the point that theyactually occurred.

Proper Application of the Paced Patient Algorithm

Careful observation during the arrhythmia systemsanalysis of the paced patient is vitally important.The clinician must always verify that paced pulsedetection is indeed taking place.

First, paced patient select must be turned on. Themonitor relearns the patients rhythm using thepaced patient algorithm.

While learning, the user should observe the delayed,annotated wave to be certain pace pulse tic marks areproperly associated with pace pulses.

Up to two pace pulse tic marks are displayedregardless whether they are associated with thefollowing beat or not. Thus for a dual-chamberpaced beat there will be two separate marks (one foreach pace pulse).

Paced Alarms

Although not designed to detect pacemaker sensingproblems, the alarm system incorporates two specificalarms for paced rhythms:

Pacer-Not-Capture is determined when a QRS

does not occur for 1.75 times the average R-to-Rinterval and a pace pulse is detected in the timeinterval.


10/16

10

Pacer-Not-Pace is initiated if a QRS does notoccur for 1.75 times the average R-to R intervaland no pace pulse is detected in the time interval.

Note: M beat label will appear when a QRSdoes not occur.

The following application points can greatlyimprove the results of the paced patient algorithm:

Pacing detection should always be turned on whenmonitoring patients with any type of pacemaker.

Only the Monitoring or Diagnostic bandwidthshould be used with paced patient monitoring.

Paced complexes should be between 1 and 2millivolts in size and taller than the pace pulse.

Ventricular paced beats should be wider than thenormal QRS complex.

Pace pulses should not have visible repolarization

(overshoot/undershoot). Repolarization causes

increased width to the pace pulse and could resultin the pace pulse being detected as a beat.

Special Concerns for ComputerizedArrhythmia Monitoring

It is impossible to design a computerized arrhythmiaalgorithm that accurately analyzes 100% of allpatients. In the following sections, several conditionsthat can cause difficulty for the algorithm aredescribed.

Tall P- and T-waves

The algorithm is designed to selectively recognizeand filter P- and T-waves to prevent classification asbeats. However, if a T-wave is much larger than theR-waves height, correct classification is difficult.The T-wave might be detected and incorrectlyclassified as a PVC, and an R-on-T PVC or HighHeart Rate alarm could be activated.

Large P-waves may also be detected and incorrectlyclassified as an R-wave, causing the algorithm togenerate incorrect high heart rate or PVC-relatedfalse alarms.

In most instances, large T- and P-waves can beaddressed by selecting different leads. However, in

conditions such as extreme atrial hypertrophy,hyperkalemia, or decreased ventricular voltage, theP- and T-waves may be as large as the R-wave despitecareful lead selection. In these cases, instead of trying

to select two leads with the proper P- and T-waveheight, it is easier to just select the lead that showsthe lowest P- and T-wave height and use the singlelead arrhythmia monitoring option.

Aberrantly Conducted Beats

Since P-waves are not analyzed, it is difficult andsometimes impossible for a monitoring system todistinguish between an aberrantly conductedsupraventricular beat and a ventricular beat. If theaberrant beat resembles a ventricular morphology, itis classified as ventricular.

Atrial Fibrillation and Flutter

Since P-wave morphology is not analyzed, there isno method to discriminate atrial rhythms. I f there isconstant variance in the R-R interval, the rhythm is

classified as Irregular.

In some cases of atrial dysrhythmias, the erraticbaseline fibrillations and flutters may be greater thanthe algorithm's detection threshold, causingerroneous detection and false alarms.This is anothercondition where single-lead arrhythmia monitoring

Large T-Waves

Atrial Hypertrophy

Aberrantly Conducted Beat


11/16

11

should be considered if it is difficult to select twoleads that have low level erratic baseline.

Intermittent Bundle Branch Block

The phenomenon of bundle branch or any of theother fascicular blocks creates a challenge for thearrhythmia algorithm. If the QRS during the blockchanges considerably from the learned normal, theblocked beat may be incorrectly classified asventricular, causing false PVC alarms.

Cardiotach Mode (Arrhythmia Analysis Off)

For the M3001A Measurement Server, the ST/ARalgorithm also provides a cardiotach function whenthe arrhythmia analysis is turned off. The cardiotach

algorithm can process one or two simultaneous ECGchannels.

For QRS detection, the cardiotach algorithm usesthe same detection signal used in the arrhythmiaalgorithm. Since all the detected beats are used in theheart rate calculation without further testing forrejection, a higher detection threshold is used toprevent the detection of low amplitude artifacts, T-waves and P-waves.

Note: When the ECG rhythm has beats withvarying R-wave amplitudes and the difference of the

R-wave is > 50%, the cardiotach may not count thelower amplitude beats. To avoid undercounting youcan select lead(s) where the beats have a similiar R-wave amplitudes or turn arrhythmia analysis on.

From the beats detected, the heart rate is thencalculated using the same formulas used in thearrhythmia algorithm. Working in parallel with beatdetection, the asystole and ventricular fibrillationdetection algorithms used in the arrhythmia analysisare also used to detect the presence of asystole andventricular fibrillation. The arrhythmia alarmsavailable are a subset of the basic arrhythmia alarms.See the note with the Alarm Chain for the BasicArrhythmia Monitoring for the alarms available.

The Arrhythmia SystemsAlarm Structure

Alarm Detection

The ST/AR arrhythmia monitoring algorithm isdesigned to analyze up to 22 rhythm disturbancesand irregularities. Each of these must pass a set of

tests before the alarm is declared. I f the system hasbeen configured to have Basic Alarms active, thesystem will declare only ten alarms. Systems thathave been configured with Enhanced Alarms willhave 22 alarms. See the charts on alarm chaining forthe alarms in each group.

After an alarm sounds and a more serious alarm isdetected, the lesser alarm message disappears and thehigher priority alarm is activated.

The following two tables describe each alarm andthe conditions required to generate the alarm.

Note: >= is greater than or equal to a number.

Atrial Fibrillation

Intermittent Bundle Branch Block

*** Red Alarms

Alarm Condition Required to

Generate Alarm

Asystole No QRS detected for xseconds. Choices of >2.5 to 4 secondsNote: M3/M4 - No QRSdetected for 4consecutive seconds

Ventricular Fibrillation/

VentricularTachycardia

Fibrillatory wave

(sinusoidal wavebetween 2-10 Hz) for 4consecutive seconds

VentricularTachycardia

Consecutive PVCs >= V-Tach Run limit and HR >V-Tach HR limit

Extreme Tachycardia Heart Rate greater thanthe Extreme Tachy limit

Extreme Bradycardia Heart Rate less than theExtreme Brady limit


12/16

12

Alarm Activation and GradedAlarm Structure

The ST/AR arrhythmia systems alarm structure isbased on priorities, with a system of *** Red,** Yellow, and INOP alarms. Each type has adistinctive visual and audible alarm, enabling quickrecognition of the severity of the alarm event.

Once an alarm is detected, it is immediatelyactivated. An alarm message appears on the display,and a distinctive audible alarm activates.

** Yellow Alarms


Generate Alarm

Non-Sustained VT A run of Vs having HR >V-Tach HR limit, butlasting for less than the V-

Tach Run limit

Ventricular Rhythm A dominant rhythm ofadjacent Vs >vent rhythmlimit and ventricular HR 1.5

to 2.5 secondsa

Note: M3/M4- No beatdetected for 1.75 xaverage R-R interval forHR 120(non-paced patient only)

Pacer Not Capture No QRS for 1.75 x theaverage R-R interval withPace Pulse (paced patient

only)

Pacer Not Pace No QRS and Pace Pulsefor 1.75 x the average R-Rinterval (paced patientonly)

Missed Beat No beat detected for 1.75x average R-R interval forHR 120(non-paced patient only)

Note: M3/M4-this alarm isnot available

SVT Run of SVPBs >= SVT Runlimit andSVT Heart Rate > SVT HRlimit

R-on-T PVC For HR Limit PVCs within one minuteexceeded the PVCs /minlimit

Multiform PVCs The occurrence of twodifferently shaped Vs, eachoccurring at least twice

within the last 300 beatsas well as each occurringat least once within thelast 60 beats

Heart Rate>Limit Heart Rate greater thanthe upper HR limit

Heart Rate 2.5 sec., if an eventoccurs of > 2.5 sec., the Asystole alarm with be annunciated.

** Yellow Alarms


Generate Alarm


13/16

13

*** Red Alarms

Red alarms, the most critical and li fe-threatening,always take priority over lesser arrhythmia alarms.They can never be individually turned off. All redand yellow alarms will be turned off if:

Alarms are suspended.

ECG alarms are off (HR alarms off ) at the

bedside. All red and yellow arrhythmia alarms are off for

telemetry.

With the M3/M4 monitor and IntelliVue MP 60/70/90 Patient Monitor, if you have configured thesystem to have the Visual and Audible Latching toOff, then the following behavior will apply when ared arrhythmia alarm occurs:

While the alarm condition is present, but has notyet been Silenced the condition will have anaudible sound and visual message.

If the condition should cease before the alarm hasbeen Silenced then the visual message and theaudible sound will cease.

If the condition is present and Silence has beenpressed, the system will maintain the visual textmessage until the condition ceases and will followthe reminder behavior that has configured.

For more detailed information on how the alarmswork refer to the M3/M4 monitor Users Manual.

** Yellow Alarms

Yellow alarms are considered lower in priority thanred alarms, but still may indicate serious rhythm orrate disturbance. A yellow alarm can be supersededby a more serious yellow alarm event, or a red alarm.

Individual yellow alarms can be disabled. Disablingyellow alarms for a particular patient does not affectany alarms on any other patient.

INOP Alarms

INOP alarms occur whenever the ECG signalcannot be properly analyzed due to noise or INOPconditions. If more than 2/3 of the time over the last30 seconds beats are classified as either noisy orquestionable, aCannot Analyze INOP alarm isgenerated. When active, the INOP alarm continues,visually and audibly, as long as the condition exists,and stops automatically when the conditionterminates.

Since the INOP alarm is a lower priority alarm itwill not override a red or yellow alarm should itoccur during the same time a red or yellow alarm isoccurring. On the other hand, if an arrhythmiaevent is detected while the INOP alarm is active, the

red or yellow arrhythmia alarm will override theINOP alarm.

Since theCannot Analyze INOP alarm indicatesthat the effectiveness of the arrhythmia monitoringfor the patient is compromised, a quick response tothis alarm is recommended.

Alarm Chaining

To prevent the confusion of redundant alarms or theactivation of less important alarms whileacknowledging serious alarms, the arrhythmiasystem sets alarm priorities through an alarmchaining system.

Related events, such as ventricular alarms, aregrouped in a chain. The most critical alarmsoccupy the top of the chain and are followed byevents in logical, descending order.

The manner in which the alarms are grouped and

prioritized define how the alarms are announced.*** Red alarms having the highest priority areannounced first if present. If there are no *** Redalarms detected, then the highest priority ** Yellowalarm detected in any given alarm chain isannounced. If alarms of the same priority indifferent alarm chains are detected, the alarm thatoccurred most recently is announced.

Alarm Chain for Basic ArrhythmiaMonitoring

Note: When arrhythmia analysis is turned off theonly available alarms are: Asystole, VentricularFibri llation, Extreme Tachycardia, ExtremeBradycardia, High HR and Low HR.

(REDALARMS)

Asystole

V-Fib/V-Tach

Extreme Tachy Extreme Brady

(YELLOWALARMS)

PNC* PNP*

V-Tach

Beat DetectionAlarms RateAlarmsPVCAlarms

First level timeout period

Secondlevel timeout period

HighHR Low HR

* PNC = Pacer Not CapturePNP = Pacer Not Pace

PVCs>xx/Min.


14/16

14

Alarm Chain for Enhanced ArrhythmiaMonitoring

Alarm Behavior after Annunciation

Alarm Timeouts

To reduce the number of unnecessary alarms, at thetime when a ** Yellow alarm is announced, a fixednon-extending timeout is initiated. The duration ofthe timeout period is user-configurable. Red alarmshave no timeouts.

During the timeout period, this alarm and all lowerpriority alarms within the same chain (group) willnot be announced. Higher priority alarms or alarmsfrom a different chain will be announced if detectedduring the timeout.

The timeout period for first level yellow alarms canbe configured for between 0 and 5 minutes. Thetimeout period for second level yellow alarms can beconfigured for between 0 and 15 minutes. Forexample:

1.If a Pair of PVCs alarm occurs,

- Yellow alarm sounds for eight seconds.

- Yellow alarm message displays for three minutesor unti l silenced.

- Timeout begins.

2.If another Pair of PVCs alarm occurs during thetimeout period,

- No alarm occurs.

- No alarm message displays.

3.If a Run of PVCs alarm (which is higher on thechain) occurs during the timeout period,

- A new yellow alarm sounds for eight seconds.


- A new timeout begins.

4.I f a Mult iform PVCs alarm (which is lower on thechain) occurs during the timeout period,

- No alarm occurs.

- No alarm message displays.

5.If aPause alarm, which is another chain, occurs

during the timeout period,

- A new yellow alarm sounds for eight seconds.


- A new timeout begins in a different chain. Thistimeout will not alter the previous timeout forthe PVC related alarm.

Timeout periods will end once the configuredamount of time has passed. They will also be cleared

if any of the conditions that cause learning occur.(see Learning)

Alarm Reminder

After a *** Red alarm is silenced the same alarm willbe re-announced, called alarm reminder, if thealarm condition persists after a fixed amount oftime. When the alarm reminder is triggered, onlythe alarm sound associated with the alarm is brieflysounded. The alarm reminder will repeat every threeminutes as long as the alarm condition persists.Alarm reminder can be configured to be on or off.

(REDALARMS)

Asystole

V-Fib/V-Tach

Extreme Tachy Extreme Brady

(YELLOWALARMS)

SVT

High HR Low HR

Irregular HR

PNC* PNP* Missed

MultiformPVCs

PVCs > xx/Min.

Vent Trigeminy

Vent Bigeminy

R-on-T PVCs

Run PVCs

Pair PVCs

NonSustain VT

VentRhythm

V-Tach

Beat DetectionAlarms RateAlarmsPVC Alarms

First level timeout period

Secondlevel timeout period

* PNC = Pacer Not CapturePNP = Pacer Not Pace

Pause

Beat

Note For M3/M4: Beat Detection Alarms

PNC* PNP* Pause


15/16

15

Points to Remember About Alarms

Clinicians should acknowledge as many alarms aspossible.

If an alarm condition exists, it is always activatedunless it is turned off, there is a higher priorityalarm in effect, or the fixed timeout period is ineffect.

** Yellow alarms can individually be user-disabled. *** Red alarms never automatically reset with one

exception. If the M3/M4 Monitor or theIntell iVue MP 60/70/90 Patient Monitor isconfigured with the Visual and Audible Latchingto OFF, any Red alarm will cease once thecondition ceases.

All arrhythmia alarms are disabled if alarms aresuspended or ECG alarms are off (HR alarms off )at the bedside. All alarms can be turned off if thepatient is being monitored via telemetry, by using

theAll Red and Yellow Arrhythmia Alarms Offbutton.

How to Get the BestPerformance fromComputerized ArrhythmiaMonitoring

Steps to Better ArrhythmiaMonitoring

1. Optimize Signal Quality

- Skin preparation is especially important whenusing gel electrodes.

- Change electrodes every 24 hours. Increasedbaseline wander is the first indication thatelectrodes are dry and need to be changed.

- Support cable and electrode wires. Artifact andwander may increases if the skin under theelectrode is stretched. Taping the electrode mayreduce this if your patient is active.

2. Choose the Best Lead(s)

- Choose a lead(s) where QRS amplitude is stableand has adequate amplitude (recommendedamplitude greater than 0.5 millivolts).

- If the system can do multi-lead analysis but onlyone lead has adequate stable voltage, changearrhythmia analysis to single-lead analysis.

3. Ensure the Best QRS Complex

Size and shape of the QRS are very important forproper beat detection and classification. Use thefollowing guidelines to choose leads which

produce the best QRS morphology for analysis bythe arrhythmia system.

The Normal Beat

- R-wave is tall, not clipped or biphasic.

- T-wave is less than 1/3 the R-wave height.

- P-wave is smaller than 1/5 the R-wave height.

The Ectopic Beat

- Height is at least 1/5 the normal QRS height.

- Beat should not be clipped.

- Shape is distinctly different than the normal.

The Paced Beat

- Pace pulse not much larger than the QRScomplex.

- Ventricular paced beat is wider than the normal

QRS complex.

- Pace pulse has no visible re-polarization.

4.Adjust Alarms

Adjusting some alarms off, changing the alarmcriteria or adjusting the timeout periods will:

- Reduce the number of alarms.

- Alert the clinician to alarms specific to thepatient.

- Prevent redundant alarms for known or chronicconditions.

ConclusionComputerized arrhythmia monitoring is a tool theclinician can use to continuously monitor andevaluate the progress of patients. In order to fullymake use of this tool, it is important to understandthe computer algorithm's capabilities andlimitations.


16/16

CORPORATE HEADQUARTERS:

Philips Medical SystemsNederland B.V.Postbus 10.0005680 DA BestNetherlands

UNITED STATES:

Philips Medical SystemsCardiac and Monitoring Systems3000 Minuteman RoadAndover, MA 01810(800) 934-7372

CANADA:

Philips Medical Systems2660 Matheson Blvd. E.

Mississauga, Ontario L4W 5M2(800) 291-6743

EUROPE, MIDDLE EAST AND AFRICA:

Philips Medizinsysteme Bblingen GmbHCardiac and Monitoring SystemsHewlett-Packard Str. 271034 BblingenGermanyFax: (+49) 7031 463 1552

LATIN AMERICA HEADQUARTERS:

Philips Medical Systems5200 Blue Lagoon Drive9th FloorMiami, FL 33126(305) 267-4220

ASIA PACIFIC HEADQUARTERS:

Philips Medical Systems24F Cityplaza One1111 King's RoadTaikoo Shing, Hong Kong

(+852) 3197 7777

www.medical.philips.com

Koninklijke Philips Electronics N.V., 2001.

All rights are reserved. Reproduction in whole or in part is

prohibited without the prior written consent of the copyright

holder.

Philips Electronics North America Corporation reserves the

right to make changes in specifications or to discontinue any

product at any time without notice or obligation and will

not be liable for any consequences resulting from the use of

this publication.

Published June 30, 2002

5990-0517EN

Philips Medical Systems is part of Royal Philips Electronics

Documents

STAR arrhythmia monitoring.pdf