Medilog Darwin Clinical Application Guide Issue 1.0

Medilog Darwin Clinical Application Guide

Preface Charles Darwin published The Origin of Species in 1859 to expound his theory of evolution by means of natural selection, a theory popularly known as the survival of the fittest. Darwin's theory applies to all life, including humans, and helps to explain our place in the world. But one way its impact on individuals is changing is through the development of new drug treatments, surgical procedures, and medical devices such as contact lenses and heart pacemakers. Another specific example – and the subject of this booklet – is the early detection of cardiac arrhythmias using ambulatory electrocardiography, as pioneered by the American biophysicist Norman Holter1 one hundred years after the publication of Darwin's book. Holter monitoring, as it became known, involves a continuous recording of the ECG while the patient goes about their normal daily life. A typical recording lasts 24 hours and contains around 100,000 heartbeats, so for this much data an automatic analysis system is essential. The analysis system scans the ECG for arrhythmias and brings them to the clinician's attention. Commercial Holter analysis systems first appeared in the early 1970s, and today's systems take the form of proprietary software running on a standard computer. The analysis algorithms used in these systems need to be fast and accurate, because ECGs are often noisy and mistakes made by the analysis algorithms are time consuming for users to correct by manual editing. But most Holter companies today are still using analysis algorithms that have not changed greatly since the 1970s. This is despite the fact that these same three decades have seen major advances in signal processing and pattern recognition techniques by university research laboratories around the world. This has led to a significant technology gap between Holter companies and universities. Medilog Darwin Holter Analysis aims to close this gap using new concepts and technologies that offer users accuracy, speed, and ease of use. Darwin's theory of evolution started a revolution in the way people see the world. Medilog Darwin represents the evolution of Holter analysis, and signals the start of a revolution in the way Holter analysis is performed. This booklet provides some background to the new concepts and technologies used in Medilog Darwin and the clinical benefits they offer. I hope you will find it interesting reading.

Dr J Pardey. London, UK, April 2005.

Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

Page 1

Medilog Darwin Clinical Application Guide

Introduction

Medilog ADAPT technology

The purpose of this Clinical Application Guide is to provide some background to the new concepts and technologies used in Medilog Darwin Holter Analysis and the clinical benefits they offer. It does not cover concepts that are already familiar to users – such as template editing, HRV analysis, ST trends, and pacemaker analysis – nor does it explain how to use Medilog Darwin, as this is all covered in the Medilog Darwin User Manual. The new concepts and technologies covered in this booklet are presented in the following order:

Medilog Darwin contains a powerful new three channel, patient adaptive, ECG analysis algorithm called Medilog ADAPT. The algorithm uses sophisticated signal processing and pattern recognition techniques, including artificial neural networks. Medilog ADAPT was presented at the 2004 International Conference on Computers in Cardiology2 and is one of the first commercial applications to contain neural networks. A brief overview of Medilog ADAPT is provided below. This first considers the importance of good hook up, and then looks at how Medilog ADAPT learns the patient's dominant beat morphology and analyses the ECG. The accuracy of Medilog ADAPT is then compared against algorithms used in other Holter systems.

-

Medilog ADAPT™ technology  ECHOView Atrial analysis Proscan Derived 12-lead ECG Apnoea screening Circadian index

Why use Medilog Darwin? Extensive customer surveys and focus groups conducted prior to the development of Medilog Darwin revealed that the top three requirements for users of a Holter system are accuracy, speed, and ease of use. Medilog Darwin Holter Analysis has been specifically designed to meet these requirements.

Patient hook up It is well known that ECG signals are susceptible to noise and artefact that degrade the ECG and make it more difficult to analyse. Common sources of noise are 50 or 60 Hertz interference from fluorescent lighting and mains appliances, and electromyographic (EMG) noise from a patient's skeletal muscles, as shown in Figure 1. Analysis algorithms can reduce or in some cases eliminate the effects of these two noise sources, but have much greater difficulties with movement artefact caused by poor electrical contact between the electrodes and the patient's skin, as shown in Figure 2. The presence of noise and artefact in the ECG can be minimised by using

Accuracy Medilog Darwin is powered by Medilog ADAPT technology – a new three channel, patient adaptive, ECG analysis algorithm that uses advanced signal processing and pattern recognition techniques. Medilog ADAPT is very accurate.

Speed Medilog Darwin takes about 75 seconds to analyse a 24-hour three channel ECG on a standard computer running Microsoft Windows XP.

Figure 1: ECG contaminated by EMG noise.

Ease of use Medilog Darwin is easy to use. Each recording is automatically analysed as it is uploaded, and new features such as ECHOView and Proscan greatly simplify the subsequent review and processing of a recording. This is in addition to more familiar features such as template editing, HRV analysis, ST trends, and pacemaker analysis.

Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

Figure 2: ECG corrupted by movement artefact.

Page 2

Medilog Darwin Clinical Application Guide

Analysing the ECG The automated analysis of a Holter recording can be divided into three sequential activities: -

Poor hook up yields poor quality recordings.

high quality ECG electrodes, and by proper skin preparation during hook up. The American College of Cardiology and American Heart Association best practice guidelines for ambulatory electrocardiography3 recommend that skin areas where electrodes are attached should be clean, dry, and shaved if necessary to remove body hair. The skin should be wiped with alcohol and gently abraded to remove dead skin. The guidelines recommend that when electrodes are attached an impedance meter is used to check that the reading between any two electrodes is below 5kΩ, and preferably below 2kΩ.

Learning the dominant morphology When an ECG is uploaded to Medilog Darwin for analysis, Medilog ADAPT uses the first few minutes of the recording to learn the patient's dominant beat morphology, which may be very different to any other ECG it has analysed. This learning period is invisible to the user as Medilog ADAPT learns very quickly and then rewinds to the start and analyses the entire recording. To facilitate learning the start of the recording should not be too noisy, and a reasonable number of beats should reflect the dominant morphology. If this is not the case – for example, if the ECG is very noisy – the user may choose to exclude the noisy region and reanalyse. Medilog ADAPT will then start its learning period at the end of the excluded region where the user has deemed the ECG to be of better quality.

Beat detection. Beat classification. Rhythm analysis.

The purpose of beat detection is to locate each heartbeat in the ECG and mark it with a beat time, also known as a fiducial or trigger point. Beat classification then uses these beat times to analyse the characteristics of each beat and classify it as a normal beat (N) or a ventricular ectopic beat (V). Rhythm analysis examines the resulting sequence of beat times and beat types to identify any arrhythmias that are present. Thus a repeated sequence of beats in the order, N V N V N V, would be identified as bigeminy. Medilog ADAPT performs beat detection and beat classification using the modular approach shown in Figure 3. Beat times are generated at the outputs of the third module, and beat types at the outputs of the fourth module. It is this fourth module that contains artificial neural networks. These neural networks learn the characteristics of each patient's ECG on each channel and continue to learn as the analysis proceeds. This patient adaptive learning makes Medilog ADAPT significantly more accurate.

ECG1 ECG2 ECG3

resampling

filtering

beat detection

beat classification

conflict resolution

The dominant beat morphology may change over the duration of a recording so Medilog ADAPT updates it as the analysis proceeds. This patient adaptive approach makes Medilog ADAPT significantly more accurate.

Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

N,N,V,N,V,N,V,…

Figure 3: The Medilog ADAPT analysis algorithm

Page 3

Medilog Darwin Clinical Application Guide Note that the first four modules in Figure 3 analyse each of the three ECG channels independently. The beat times and beat types from all three channels are then compared to yield a single sequence of beat times and beat types for subsequent rhythm analysis. Noise or artefact on one or more channels may mean that the beat times and beat types on each channel do not always agree, so Medilog ADAPT contains a number of innovations to deal with any differences. These innovations are used to accept or reject a beat that has been detected on only a subset of the channels, and to arrive at a final beat type when there are differences on each channel. To do this Medilog ADAPT monitors the signal quality on each channel and assigns a high, medium, or low priority to each channel accordingly. Good quality channels receive a high priority while channels with noise or artefact receive a low priority. This ensures that results from good quality channels are given a greater weighting than results from poor quality channels. The channel priorities may change as the analysis proceeds to reflect changes in the quality of the recording, such as occasional periods of noise or artefact on an otherwise clean channel. This dynamic switching of channel priorities enables Medilog ADAPT to use good quality data whenever they are available. Medilog ADAPT can even continue to work during periods when two channels are unusable due to signal drop-out or noise corruption. The use of three channel analysis therefore improves the accuracy and robustness of Medilog ADAPT beyond that of two channel algorithms.

Superior performance The accuracy of an analysis algorithm can be assessed by measuring how good it is at QRS detection and ventricular ectopic beat (VEB) detection. For example, a QRS complex that is correctly detected is counted in the algorithm's favour as a true positive (TP), but a missed QRS complex is counted against the algorithm as a false negative (FN) and a falsely detected QRS complex where there isn't one is counted against the algorithm as a false positive (FP). The number of true positives, false positives and false negatives can then be used to calculate two performance measures called sensitivity (Se) and positive predictivity (+P), as follows:

Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

Why use neural networks? Neural networks contain artificial neurons called perceptrons that mimic the behaviour of biological neurons in the brain. A neural network contains multiple layers of interconnected perceptrons and can learn to recognise complex patterns. The use of neural networks in Medilog ADAPT originates from some earlier research conducted at Oxford University in which neural networks were applied to the analysis of human electroencephalography (EEG) signals recorded during sleep4. The conventional approach to analysing sleep is to divide the EEG into 30-second blocks and assign each block to one of six sleep stages using scoring rules published in 1968. The Oxford University research demonstrated for the first time that neural networks can learn to track the waxing and waning of sleep as a continuous biological process, thereby revealing the microstructure of sleep that is missed by conventional sleep staging. This ability of neural networks to learn complex and often subtle patterns in physiological signals makes them ideally suited to ECG analysis, as described in this Clinical Application Guide.

Se = TP / (TP + FN) +P = TP / (TP + FP) A good analysis algorithm will have high sensitivity and high positive predictivity, but this is often difficult to achieve and there is usually a trade off between the two measures. For example, an algorithm can achieve 100% sensitivity (no false negatives) by labelling everything as a QRS complex but if it does this there will be many false positives and its positive predictivity will be close to 0%. These performance measures can also be calculated for VEB detection but the same trade off applies – if an algorithm labels everything as a VEB its sensitivity will be 100% (no false negatives) but its positive predictivity will again be close to 0%. The EC38 and EC57 standards5 published by the American National Standards Institute and the Association for the Advancement of Medical Instrumentation require the sensitivity and positive predictivity of ECG analysis algorithms

Page 4

Medilog Darwin Clinical Application Guide to be reported for both QRS and VEB detection. The standards state that these tests should be performed without user intervention on two reference databases of digitised two channel Holter recordings. These are the American Heart Association (AHA) Database for Evaluation of Ventricular Arrhythmia Detectors, and the Massachusetts Institute of Technology and Beth Israel Hospital (MIT-BIH) Arrhythmia Database. The AHA database contains 78 non-paced ECGs, each of 35 minutes duration, and the MIT-BIH database contains 44 non-paced ECGs, each of 30 minutes duration. Medilog ADAPT has been tested on these ECGs in strict accordance with the EC38 and EC57 standards. The gross statistics for QRS and VEB sensitivity and positive predictivity are reported as percentages in the tables below, along with published results for the Spacelabs Multiview algorithm6, the Philips ST/AR algorithm7, and the Mortara Veritas algorithm8. The results show that Medilog ADAPT outperforms 19 of the 24 performance measures reported for the other three analysis algorithms.

Multiview ST/AR Veritas ADAPT

QRS Se QRS +P 99.76 99.87 99.80 99.87 99.88 99.89 99.90 99.90

VEB Se 95.44 95.96 94.07 98.42

VEB +P 96.99 98.34 97.72 97.62

Performance results on the AHA database.

Multiview ST/AR Veritas ADAPT

QRS Se QRS +P 99.82 99.85 99.66 99.86 99.93 99.85 99.87 99.91

VEB Se 93.00 94.25 95.44 97.46

VEB +P 90.05 96.38 96.90 93.60

Performance results on the MIT-BIH database.

ECHOView ECHOView is a powerful new feature that is unique to Medilog Darwin. It offers users the ability to view full disclosure ECG 'at a glance' by displaying it as a colour map in which arrhythmias are visually highlighted. To generate this colour map ECHOView colour codes the amplitude of the ECG signal and uses the QRS trigger points to align the QRS complexes alongside each other. ECHOView then displays the colour coded and time aligned QRS complexes 'from Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

Figure 4: Medilog Darwin's ECHOView screen. above' so that each is now a single vertical line of varying colour. The juxtaposition of these lines from left to right across the screen forms the ECHOView colour map. A typical Medilog Darwin ECHOView screen is shown in Figure 4. The upper window is the ECHOView colour map. This shows normal sinus rhythm, in blue, and two brightly coloured periods of ventricular tachycardia. The rhythm disturbances are immediately obvious to the user, who can then confirm them by positioning the cursor in the ECHOView window. This causes the corresponding section of ECG to be displayed in the lower window, as shown in Figure 4.

Atrial analysis Medilog Darwin atrial analysis facilitates the rapid detection of atrioventricular (AV) blocks, atrial flutter and atrial fibrillation. Medilog Darwin atrial analysis works by first dividing the ECG into short, typically five minute sections and generating a histogram of the PR intervals in each section. These histograms are then colour coded and displayed from above as juxtaposed vertical lines. A typical Medilog Darwin atrial analysis screen is shown in Figure 5. In this example the colour coded histograms in the upper window form a thin line that is frequently above the 200 millisecond limit for normal PR intervals, thus revealing the presence of firstdegree AV block. Towards the right hand side of the upper window the line disperses into an area of colour that signifies a period of atrial fibrillation. Positioning the cursor in this area displays the corresponding five minute histoPage 5

Medilog Darwin Clinical Application Guide

Figure 5: Medilog Darwin's atrial analysis screen.

Figure 6: Medilog Darwin's Proscan screen.

gram in the middle window. This histogram has a wide spread of PR intervals, as expected, and atrial fibrillation is confirmed by the corresponding section of ECG in the lower window. For a healthy patient with constant PR intervals the colour coded histograms in the upper window would appear as a straight line with a narrow spread of PR intervals in the middle window. Medilog Darwin atrial analysis is available for recordings made using a Medilog AR4 or AR12 Holter recorder.

a special vectorcardiography (VCG) patient cable to produce the three orthogonal XYZ leads from which the twelve standard leads – I, II, III, aVR, aVL, aVF and V1 to V6 – are then derived using a set of mathematical equations9. A typical Medilog Darwin derived 12-lead ECG screen is shown in Figure 7. The upper right hand window contains a three dimensional plot of the XYZ vector loop (which the user can rotate by dragging it with the mouse) and three twodimensional projections of the vector loop onto each orthogonal lead pair, where XY is the frontal view in green, XZ is the horizontal view in blue, and YZ is the sagittal view in red. The derived 12-lead ECG itself is displayed in the lower window, and two of these derived leads – I and II in this example – are plotted relative to each other as a Cabrera circle in the upper left hand window. The magnitude and angle of the heart vector are displayed in the lower left hand corner of the Cabrera circle.

Proscan Proscan is a new prospective scanning tool that performs high speed playback of full disclosure ECG and stops whenever one of several predefined arrhythmia limits is exceeded. Examples of these arrhythmia limits are VEBs per minute, pauses, and upper or lower heart rate. A typical Medilog Darwin Proscan screen is shown in Figure 6. The arrhythmia limits are set using cursor controlled sliders in the upper right hand window and the high speed ECG playback is displayed in the upper left hand window. When Proscan stops at an arrhythmia the corresponding section of ECG is displayed in the lower window. The user can then confirm the arrhythmia and continue playback from that point.

Derived 12-lead ECG Medilog Darwin derived 12-lead ECG offers continuous 12-lead ECG monitoring from only five electrodes. The electrodes are connected to

Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

Figure 7: Medilog Darwin's derived 12-lead ECG.

Page 6

Medilog Darwin Clinical Application Guide Note that a 12-lead ECG derived from five electrodes is an approximation to a conventional 12lead ECG obtained from ten electrodes and should not be used for diagnostic purposes. Medilog Darwin derived 12-lead ECG is available for recordings made using a Medilog AR12 Holter recorder and VCG patient cable using the following RAFE electrode configuration10: R = Upper right arm. A = Left fifth intercostal space, mid-axillary line. F = Upper left leg. E = Anterior median line, horizontal to A. A fifth electrode on the right eighth rib is used to balance the potential between the patient and the recorder. Electrode configurations other than RAFE will produce incorrect results.

Apnoea screening An apnoea is a sleep-related breathing disorder in which the sufferer stops breathing during sleep. An apnoea is terminated by an arousal from sleep and a period of hyperventilation, followed by an eventual return to sleep. This may be repeated several hundred times during a single night, but because the arousals are not to full wakefulness sufferers are often unaware of the condition. There are two types of apnoea: Central sleep apnoea (CSA) originates in the brainstem and leads to a loss of respiratory effort, i.e. the brain 'forgets' to tell the body to breathe. The incidence of CSA is small. Obstructive sleep apnoea (OSA) is caused by the upper airway collapsing which prevents the patient from breathing. Respiratory effort is sustained in the diaphragm and chest wall muscles but the airway itself is physically blocked. OSA is common in obese patients. Chronic apnoea makes proper sleep impossible, resulting in excessive daytime sleepiness and impaired cognitive function. Chronic apnoea has been linked with road traffic accidents as well as increased risk of stroke and heart failure, so its detection is important. Medilog Darwin performs apnoea screening directly from the ECG without the need for additional equipment. The method is based on respiratory sinus arrhythmia (RSA), a well known phenomenon in which the

Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

Figure 8: Medilog Darwin's apnoea screen. heart rate speeds up and slows down as the patient breathes. RSA can be observed in the ECG as a progressive shortening and lengthening of the RR intervals. In addition the QRS amplitudes may also increase and decrease due to slight changes in the position of the ECG electrodes relative to the patient’s heart as the lungs empty and fill. Medilog Darwin tracks these changes to provide information about the patient's breathing. Temporary losses of RR interval or QRS amplitude changes are highlighted as potential apnoeas11. A typical Medilog Darwin apnoea screen is shown in Figure 8. The upper right hand window is an apnoea overview, and potential apnoeas are shown in the upper left hand window. The two lower windows show the cyclical variation in heart rate above the derived respiration signal in which repeated episodes of apnoea and hyperventilation are clearly evident. Medilog Darwin apnoea screening is available for recordings made using a Medilog AR12 Holter recorder.

Circadian index Medilog Darwin HRV analysis includes a new parameter called the circadian index (CI). This is the ratio of daytime to night time mean heart rates, where daytime is defined as 7am–10pm and night time is 11pm–6am. A study of over 7,500 healthy subjects revealed that normal CI is 1.33±0.05 irrespective of age or gender. Abnormal CI is indicated in a range of diseases, including heart failure and diabetes mellitus12.

Page 7

Medilog Darwin Clinical Application Guide

Glossary

References

Apnoea A temporary cessation of breathing. Artificial neural network A pattern recognition technique comprising multiple layers of interconnected perceptrons that can learn to recognise complex and often subtle patterns. Cabrera circle A two-dimensional plot of the heart vector as seen from two of the leads in a 12-lead ECG. Circadian index The ratio of daytime to night time mean heart rates where daytime is defined as 7am–10pm and night time is 11pm–6am. Derived 12-lead ECG An approximation to a conventional 12-lead ECG derived from fewer than ten electrodes. ECHOView A powerful new feature that is unique to Medilog Darwin and offers users a new way of viewing full disclosure ECG. ECHOView displays the ECG as a colour map so that arrhythmias may be visually highlighted. False negative (FN) A QRS or VEB that is missed by an analysis algorithm. False positive (FP) A QRS or VEB detected by an analysis algorithm when no QRS or VEB is present in the ECG. False positives are caused by noise or artefact. Impedance The electrical resistance between an ECG electrode and the patient's skin; it should be below 5kΩ and preferably below 2kΩ. Medilog ADAPT A new three channel, patient adaptive, ECG analysis algorithm that uses advanced signal processing and pattern recognition techniques, including artificial neural networks. Medilog Darwin Holter Analysis The new Holter analysis described in this booklet, powered by Medilog ADAPT technology. Perceptrons Artifical neurons that mimic the behaviour of biological neurons in the brain. Positive predictivity (+P) The ratio of true positives to true positives and false positives. Proscan A prospective scanning tool that performs high speed playback of full disclosure ECG and stops at pre-defined arrhythmia limits. RAFE An electrode configuration used in vectorcardiography that requires only five electrodes. Sensitivity (Se) The ratio of true positives to true positives and false negatives. True positive (TP) A QRS or VEB that is correctly detected by an analysis algorithm. Vectorcardiography (VCG) A three channel ECG comprising frontal, horizontal, and sagittal views of the heart that are mutually orthogonal.

1

Issue 1.0 (c) 2005 Huntleigh Healthcare Ltd

Holter NJ. "New method for heart studies: continuous electrocardiography of active subjects over long periods is now practical." Science, 1961, 134:1214-20.

2

Pardey J, Jouravleva S. "The next generation Holter revolution: from analyse-edit-print to analyse-print." IEEE International Conference on Computers in Cardiology, 2004, 31:373–6.

3

Crawford MH, Gibbons RJ, et al. "ACC/AHA guidelines for ambulatory electrocardiography." Journal of the American College of Cardiology, 1999, 34:912-48. 4

Pardey J, Roberts S, Tarassenko L, Stradling J. "A new approach to the analysis of the human sleep/wakefulness continuum." Journal of Sleep Research, 1996, 5:201-10.

5

ANSI/AAMI EC38:1998 Ambulatory electrocardiographs; ANSI/AAMI EC57:1998 Testing and reporting performance results of cardiac rhythm and ST-segment measurement algorithms.

6 Spacelabs Medical. "Arrhythmia system perfor-

mance evaluation guidelines." Spacelabs Clinical Bulletin Series, 1996; 6(11). 7

Philips Medical Systems. "ST segment and arrhythmia monitoring." Application Note 4522982-91531, 2004.

8

Mortara Instruments. "Arrhythmia algorithm performance analysis." www.mortara.com/OEM _arrhythmiadetect_specs.htm

9 Dower GE,

Machado HB, et al. "On deriving the electrocardiogram from vectorcardiographic leads." Clinical Cardiology, 1980, 3:87-95.

10 Dower GE, Osborne JA "A clinical comparison

of three VCG lead systems using resistancecombining networks." American Heart Journal, 1958, 55:523-34. 11

Penzel T, McNames J, et al. "Systematic comparison of different algorithms for apnoea detection based on electrocardiogram recordings." Medical & Biological Engineering & Computing, 2002, 40:402-7. 12 Makarov LM. "Circadian index as indicator of

stable organization of heart circadian rhythm." Clinical Medicine, 2000, 78:24-7. Page 8

www.medilogholter.com

Belgium Groene Hofstraat 31 2850 Boom, Belgium Tel: +32 (0)3 880 82 80 Fax: +32 (0)3 844 78 42 E-Mail: sales-be@oxinst.be China Room 714, Office Tower 3, Henderson Center, No 18 Jianguomennei Ave Dongcheng District Beijing 100005, China Tel: +(8610) 6518 8160/1/2 Fax: +(8610) 6518 8155 E-Mail: info@oxford-instruments.com.cn France Novamed, 16 Rue du 8 Mai 1945 -ZA - BP 47 -, 91470 Limours, France Tel: +33 (1) 64 91 24 94 Fax: +33 (1) 64 91 55 69 E-Mail: info@novamed.fr Germany Postfach 4509 65035 Wiesbaden, Germany Tel: +49 (0)6122 937-0 Fax: +49 (0)6122 937-100 E-Mail: msd.sales@oxford.de Italy Via Leone Tolstoi, 86 20098 San Giuliano Milanese Milan, Italy Tel: +39 02 982 531 Fax: +39 02 982 41407 E-Mail: sales@oxinst.it Japan Haseman Building, 2-11-6 Tomioka Koto-ku, Tokyo 135-0047, Japan Tel: +81 (0)3 5245 3591 Fax: +81 (0)3 5245 4466 E-Mail: sales.medical@oxinst.co.uk Netherlands Avelingen West 1 4202 MS Gorinchem, Netherlands Tel: +31 (0)183 692111 Fax: +31 (0)183 692119 E-Mail: sales-nl@oxinst.be Singapore 371 Beach Road, Keypoint #02-07, Singapore 199597 Tel: +65 6337 6848 Fax: +65 6337 6286 E-Mail: oimed.sales@oxford-instruments.com.sg Spain Avda. Mata Piñonera 2 28700 San Sebastian de los Reyes Madrid, Spain Tel: +34 (91) 653 8198 / 8387 / 8398 Fax: +34 (91) 654 6794 E-Mail: oxinst.msd@oxinst.es Medilog Darwin Holter Analysis complies with the EN 60601-1 safety standard. The CE mark indicates that it meets the essential requirements of European Council Directive 93/42/EEC of 14 June 1993 concerning medical devices. Huntleigh Healthcare operates a quality system certified to ISO 9001:2000, ISO 13485:2003, and the TickIT Guide Issue 5. Medilog is a registered trademark of Huntleigh Healthcare. ECHOView is a registered trademark of TOM Medical Entwicklungs GmbH. Microsoft and Windows are registered trademarks of Microsoft Corporation. For the USA the following applies: CAUTION: Federal law restricts this device to sale by or on the order of a physician. Information in this document is subject to change without notice. No part of this document may be reproduced or transmitted in any form or by any means, electronic or mechanical, for any purpose, without the prior written permission of Huntleigh Healthcare. The greatest care has been taken in the compiling of this document. However, the possibility of errors cannot be ruled out. The publisher, author, and translator accept no liability for errors or omissions and their consequences. Copyright © 2005 Huntleigh Healthcare. All rights reserved.

Part Number: 056W008–A

Huntleigh Healthcare, Diagnostic Products Division, 35 Portmanmoor Road, Cardiff CF24 5HN, United Kingdom. Tel: +44 (0)29 204 85885, Fax: +44 (0)29 204 56433.

0088

United Kingdom ScanMed Medical, Unit 4/5 Draycott Business Centre, Moreton in Marsh Gloucestershire GL56 9JY, UK Tel: +44 (0)1386 701564 Fax: +44 (0)1386 701800 E-Mail: oxford@scanmed.co.uk United States of America Advanced Biosensor Inc 6 Woodcross Drive Columbia South Carolina 29212, USA Tel: 803 407 3044 Toll Free: 800 443 3816 Fax: 803 407 3046 E-Mail: sales@advancedbiosensor.com