Available online at www.sciencedirect.com

y 42 (2009) 580–583www.jecgonline.com

Journal of Electrocardiolog

Improving sensing and detection performance in subcutaneous monitorsPeter van Dam,a Chris van Groeningen,b Richard P.M. Houben,b David R. Hampton, PhD⁎,b

aDonders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center, Nijmegen, The NetherlandsbMedtronic Bakken Research Center, Maastricht, The Netherlands

Received 28 May 2009

Abstract Implantable loop recorders (ILRs) are used for continuous assessment of patients at risk for syncope

⁎ Corresponding aThe Netherlands.

E-mail address: d

0022-0736/$ – see frodoi:10.1016/j.jelectroc

and arrhythmia. Device accuracy depends on appropriate sensing of the patient's electrocardiogram(ECG) signal. However, current methods for sensing cardiac electrical activity rely on simplethreshold detectors that are computationally efficient but nonspecific.We test the hypothesis that better ILR implant positions will increase detection accuracy. Ten healthysubjects were studied as they assumed 12 different postures. Body surface potential map (BSM)recordings were used to estimate bipolar R-wave amplitudes for 64 potential implant sites at 360orientations per site. Optimal sites were identified as the combination of position and orientation thatconsistently gave the largest signal and the lowest variability during posture changes.Results showed that posture impacts the R-wave amplitude in both BSM and derived bipolar ECGsin healthy subjects. Specific postures are associated with significant drops in R-wave signalamplitude that could cause loss of signal detection in ILRs, especially in positions likely to displacethe diaphragm. R-wave changes occurred abruptly as posture was changed. Optimal implantlocations cluster near the center of the chest, aligned with the cardiac axis, consistent with the steeperisoelectric gradients known to be associated with these positions.© 2009 Elsevier Inc. All rights reserved.

Keywords: Subcutaneous monitoring; R-wave amplitude; Detection accuracy; Sensing; Detection

Implantable loop recorders (ILRs) have become standardclinical tools for diagnosis of unexplained syncope andmonitoring for infrequent atrial arrhythmias.1,2 Reveal (Med-tronic, Minneapolis, MN), a typical ILR device, acquires abipolar electrocardiogram (ECG) signal using subcutaneouselectrodes spaced 4.5 cm apart, sensing sequential QRScomplexes, and processing the derived R-R interval series todetect pauses and irregularities in the patient's cardiac rhythm.If significant events occur, it stores relevant ECG segments forlater download and review; data can also be captured by patientactivation when characteristic symptoms are recognized. Anupdated guideline for routine clinical use of ILRs was recentlypublished by the European Society of Cardiology.3

As clinical experience with these instruments grows, sodoes an appreciation of their dependence on appropriatesensing and detection of the patient's ECG signal. Effectivememory management and efficient clinical review requirethat sensing is highly specific for R waves and that detectionalgorithms rule out episodes of muscle noise or bigeminy. In

uthor. Medtronic Bakken Research Center, Maastricht,

avid.r.hampton@live.com

nt matter © 2009 Elsevier Inc. All rights reserved.ard.2009.06.024

particular, undersensing of normal QRS complexes due toamplitude or morphology shifts may lead to false asystoledetections. Oversensing of large T waves, paced beats, orpremature ventricular contractions may lead to false reportingof irregularities suggestive of atrial fibrillation or tachycardia.

Characterization of inappropriately detected episodes4 hasmotivated improvements in both sensing methods anddetection algorithms. A retrospective analysis of 2613previously recorded automatically detected episodes from533 patients monitored by the older Reveal Plus ILRdetermined that 71.9% of episodes were inappropriatelydetected and at least 88.6% of patients had one or moreinappropriate episodes. New sensing criteria, using adaptiveR-wave sense thresholds and enhanced noise rejectiontechniques, reduced inappropriate detections by 85.2% withonly a small 1.7% reduction in the detection of appropriateepisodes. A recent prospective trial of 60 patients5 comparingReveal XT's enhanced detection sensitivity for atrialfibrillation against Holter recordings demonstrated corre-spondence of 90.6% for episodes longer than 2 minutes and amean positive predictive value of 55.1%. False-positivedetections clustered within a few patients having persistentirregular rhythms unrelated to atrial fibrillation.

Fig. 1. Positions of body surface map recording leads: positions 1 to 64 were used for analysis; locations of anatomical landmarks and standard precordialpositions are indicated.

581P. van Dam et al. / Journal of Electrocardiology 42 (2009) 580–583

Long-term patient monitoring with ILRs can, however,generate significant numbers of false-positive events evenwhen error rates are low, potentially eroding user confidence.Additional gains in performance may be achieved byenhancing the ECG signal quality through better deviceplacement or improved electrode design. Signal amplitude,for example, is dependent on heart position and orientation,which changes during normal activity of ambulatorypatients.6 The current recommended implant site in the leftshoulder may not be optimal for monitoring bipolar ECGsignals if the heart moves in the torso during posturalchanges. In addition, an ILR implanted at this site may bepulled away from the heart during movements of the arm (eg,reaching) or toward the heart during compression of theshoulder (eg, lying on that side).

The objective of this study is to investigate the relationbetween posture changes and changes in the amplitude of theR wave in normal human subjects for a range of potentialimplant sites. Locations and orientations for which signal

Fig. 2. A, Creation of R-wave amplitude graphs. Shown are 4 bipolar reconstructio180°, and 270° orientations. At every orientation, the amplitude of the bipolar R wshowing superimposed results for 10 subjects. The light gray area in the backgroun

amplitude is large and minimally influenced by changes inposture will be identified to suggest positions with potentialto enhance sensing performance in current ILR devices byminimizing posture-related variation in signal amplitude.

Methods

The pilot clinical research project is an observationalstudy of a convenience sample; no statistical analysis wasperformed due to the limited sample size. Ten healthyvolunteers with no cardiac history were enrolled; their meanage was 42 ± 6 years, and 7 subjects were male. Subjectshad normal Dutch body habitus: height, 179 ± 10 cm;weight, 78 ± 13 kg; chest circumference, 98 ± 8 cm; andsternal length, 21 ± 3 cm.

Data collection

Body surface potential map ECGs (BSM) were recordedfrom each subject using the 96-lead system illustrated in

ns derived from recordings of a single beat, with estimations of the 0°, 90°ave is estimated. B, Resulting R-wave amplitude graph for 360 steps of 1°d indicates 0.75 mVamplitude; the dark gray area, an amplitude of 0.1 mV

,,

.

582 P. van Dam et al. / Journal of Electrocardiology 42 (2009) 580–583

Fig. 1. The circumference of the chest was measured at thelevel of V2 and divided into 18 equal distances. The BSMrecording strips, each holding 8 recording electrodes, wereplaced on the subject's torso at 12 of these equidistantlocations, with the fourth electrode positioned at the level ofV2 and strips 5 and 6 positioned to the right and left of thesternum. A standard signal reference was derived fromadditional electrodes placed at VR, VL, and VF.

Subjects were asked to assume a series of fixed posturesfor 30 seconds each, whereas the resting ECG was

Fig. 3. Optimal implant location based on pooled results across subjects andembedded arcs show the number of subjects with bipolar R-wave amplitudesimplant locations and orientations, where the amplitude is less than 0.05 mVorientations for which the amplitude is at least 0.05 mV but less than 0.1 mwhere the R-wave amplitude remains above the 0.1 mV for all subjects and

recorded from the BSM array. These 12 postures comprisethe following:

• supine, lying on the back;• prone, lying on the stomach;• lying on left and right sides, each in a relaxed and in astiff position;

• standing in 4 positions: with arms hanging loosely,with the left arm up, with the right arm up, and withboth arms up;

postures. Each circle shows the net result for a BSM array locationgreater than 0.1 mV at a corresponding orientation. Red indicates pooin at least one of the subjects; yellow indicates implant locations andV in at least one of the subjects; and green highlights indicate zonepostures.

;r

s

583P. van Dam et al. / Journal of Electrocardiology 42 (2009) 580–583

• standing, with torso bent over; and• crouching, with torso erect.

Data analysis

Because subcutaneous recording devices are implanted inthe chest, only BSM leads 1 to 64 were used forreconstruction of the BSM. Representative beats wereselected from each recording interval based on consistentRMS amplitude and minimal muscle artifact across allchannels. However, because most data collection wasperformed in resting positions, few segments were omitteddue to muscle artifact. An average beat was constructed usingthe selected beats from each recording interval, and the peakR-wave amplitude was automatically determined after base-line correction. Bipolar R-wave amplitudes were estimatedby computing the difference between measured R-waveamplitudes and the interpolated value 4.5 cm away. The effectof position was assessed by repeating the calculation for eachof the 64 measured locations of the BSM array; the effect oforientation was assessed by interpolation of the BSM signalsat 360 angular increments at each BSM location, as shown inFig. 2A. A polar graph of bipolar R-wave amplitude couldthus be reconstructed for each electrode location andorientation for each subject, as shown in Fig. 2B.

Results

The estimated bipolar R-wave amplitude varied as patientsassumed each of the 12 specified postures. The degree ofchange, defined using the R-wave amplitude plots, dependedon the location and orientation of the bipolar recordingelectrodes. The implant positions with the smallest associatedchanges with posture were seen adjacent to the sternum in thevicinity of conventional leads V2 and V3. Orientationsparallel to the R-wave electrical axis gave the largest signalsand the best overall immunity to changes in amplitude. Moredistant locations had lower bipolar R-wave amplitudes andgreater proportionate variation during posture changes.

Bipolar R-wave amplitude changed immediately uponposture change, confirming the likely dependence of theeffect on anatomical shift of the heart. Crouching posturesand those with both arms raised consistently associated withthe most pronounced changes in amplitude. In contrast, lyingon the left side, expected to cause a large change in R-waveamplitude, did not show the predicted effect.

Observed variation in bipolar R-wave amplitude acrossthe 12 postures are summarized in Fig. 3. Green areasindicate locations and orientations with no sensing problemsat any posture in any of the subjects.

Discussion

The study results show a clear impact of posture on theBSM and on the derived bipolar ECGs in healthy subjects.These changes are large enough to cause significant drops inR-wave signal amplitude that could cause loss of signaldetection in ILRs. Transitions from large R waves to small

ones occur abruptly, in as few as 2 consecutive beats, assubjects change their position, consistent with the hypothesisthat changes driven by anatomical shifts in the heart'sposition. Postures that displace the diaphragm, such ascrouching and reaching, are consistently associated with themost pronounced changes in R-wave amplitude.

Although a small number of subjects were studied and therange of torso size is limited to the normal Dutch habitus, thepotential effects of age, sex, and torso size were assessed byvisual comparison of R-wave amplitude plots. Green et al7

(1985) has reported that body surface potential patternsremained constant to age 40 and that there was a tendencytoward a more horizontal zero potential line in slendersubjects. We also found no visible differences in results whengrouped by age or torso size. In female subjects, peak bipolarR-wave amplitudes were found in lower BSM positions thanin male subjects. This effect is likely to be a result of thebreast contours, which cause lower electrodes to elevate tosomewhat higher positions on the torso than in men whenstrip electrodes are used. No other sex-related differenceswere found.

Our preliminary data indicate that the effect of posture onsignal amplitude varies with electrode location and orienta-tion on the chest and that optimal implant sites may exist thatminimize these changes. We did not assess the effect ofelectrode spacing on the effect. It is not clear from our datawhether a universally “best” implant location exists orwhether it must be determined for each patient throughpreimplant mapping. However, locations near the center ofthe chest appear to be consistently better than moreperipheral locations, perhaps associated with the steeperisoelectric gradients known to be present at these positions.Currently, many physicians choose the standard pacemakerimplant location for ILRs, usually in the left shoulder region,but our data suggest that this is a relatively poor choice andthat many reported sensing problems may simply be relatedto the properties of this site.

References

1. Ashby DT, Cehic DA, Disney PJS, Mahar LJ, Young GD. A retrospectivecase study to assess the value of the implantable loop recorder for theinvestigation of undiagnosed syncope. Pacing Clin Electrophysiol2002;25:1200.

2. Ng E, Stafford PJ, Ng GA. Arrhythmia detection by patient and auto-activation in implantable loop recorders. J Interv Card Electrophysiol2004;10:147.

3. Brignole M, Vardas P, Hoffman E, et al. Indications for the use ofdiagnostic implantable and external ECG loop recorders. Europace2009;11:671.

4. Brignole M, Bellardine Black CL, Thomsen PEB, et al. Improvedarrhythmia detection in implantable loop recorders. J CardiovascElectrophysiol 2008;19:928.

5. Hindricks G, TaborskyM,Wohlgemuth P, Rieger G, Beckers F, Albers B.Atrial fibrillation detection by a subcutaneous monitoring device.Comput Cardiol 2008;35:413.

6. Sutherland DJ, McPherson DD, Spencer CA, Armstrong CS, HoracekBM, Montague TJ. Effects of posture and respiration on body surfaceelectrocardiogram. Am J Cardiol 1983;52:595.

7. Green LS, Lux RL, Haws CW, Williams RR, Hunt SC, Burgess MJ.Effects of age, sex, and body habitus on QRS and ST-T potential maps of1100 normal subjects. Circulation 1985;71:244.