Interpretation of Electrocardiogram Images Sent Through the Mobile Phone Multimedia Messaging Service

Abstract

Objective: In this study, the diagnostic accuracy of interpretations of electrocardiogram (ECG) images taken by a mobile phone and sent as multimedia message was investigated. Materials and Methods: The ECGs of 305 patients who were admitted to the emergency department with cardiac complaints were photographed with the camera of a Nokia (Espoo, Finland) N93 mobile phone. The images were sent via a multimedia messaging system to an identical mobile phone carried by a cardiologist and were interpreted on the screen of that mobile phone. Another cardiologist and an emergency physician interpreted ECG paper printouts separately. The findings of the core laboratory were used as the gold standard. The interpretation errors were scaled from 1 to 4 with respect to the significance of findings. Results: The total ratio of Grade 4 errors, which consisted of significant errors, did not show any significant difference (p=0.76) between the interpretations by the emergency medicine specialist and the cardiologist who interpreted the ECGs on the mobile phone; the cardiologist who interpreted the ECG paper printouts made significantly fewer mistakes than the other two specialists (p=0.025 and p=0.023, respectively). The separate assessment of the findings showed that in the diagnostic process of ST-segment elevation, depression, and supraventricular tachycardia, the consistency of the interpretations (κ=0.81, κ=0.81, and κ=1.0, respectively) made on the mobile phone screen was slightly better than that of the emergency medicine specialist (κ=0.73, κ=0.77, and κ=0.80, respectively) and was similar to that of the cardiologist (κ=0.91, κ=0.91, and κ=1.0, respectively) who interpreted ECG paper printouts. Conclusions: Our findings suggest that sending the ECG images via a multimedia message service may be a practical and inexpensive telecardiology procedure.

Introduction

Cardiovascular Emergencies, Time to Treatment, and Electrocardiogram Interpretation

In patients with cardiac symptoms, the most commonly performed diagnostic investigation, after a history and physical examination, is 12-lead electrocardiography. It is critical to reduce the time between triage and treatment of cardiac patients. The amount of time before reperfusion for patients with an acute ST-segment elevation myocardial infarction, for example, is an important predictor of outcome.¹ Accurate electrocardiogram (ECG) interpretation requires knowledge, skill, and experience. Computer interpretations of ECGs are useful adjuncts to physician interpretation. However, associated errors are still common.² To shorten delays, the ECG taken by emergency department (ED) personnel is often transmitted to a physician for evaluation.³ The ED physician often must interpret the ECG in difficult circumstances. Any mistake in the process of interpretation of the ECG by an emergency physician may result in improper patient treatment. A large study suggested that failure to interpret the ECG accurately was a significant factor in patient management errors.⁴ Several studies found a significant discordance in the interpretation of ECGs between emergency medicine residents and cardiologists. However, serious errors that may affect medical management are much rarer (3–4%).^5,6 The absence of cardiologists in the ED can delay the diagnosis and thus adequate management of cardiovascular emergencies. A telecardiology service can help to identify patients with urgent problems requiring expeditious treatment. It is difficult to establish an exact diagnosis through only a verbal report of the ECG. In these situations, the ECG image data can be sent to an on-call cardiologist.

ECG Transmission Methods

ECGs have been sent to receiving stations in hospitals since 1987.⁷ Recently, the transmission of 12-lead ECG images from remote locations to handheld devices (mobile phone or pocket computer) of cardiologists has been made possible with the development of wireless technology.⁸ The cost to upgrade to wireless capability was approximately $11,000 for a receiving station, $600 for cell phones, and $4,500 for data cables.⁹ Cost restrains the common use of such systems. Mobile phones may be alternative devices for telemedicine and replace receiving stations. Studies have shown that an ECG diagnosis can be accurately established by viewing the ECG on a liquid crystal display,^8,10 and thus the ECG can be transmitted directly to a handheld device carried by the attending physician for early diagnosis and triage.¹¹ In a small study, a novel approach using a camera phone with multimedia message service (MMS) was tested in 10 patients.¹²

MMS

Today's mobile phones have advanced applications. MMS is a standard for telephone message systems allowing the sending of multimedia objects (images, audio, video). Most mobile phones support the service. The most important application area for multimedia messaging is mobile-to-mobile messaging. In the field of medicine, MMS use is limited, although it takes only a few minutes to send and receive a multimedia message, and it allows senior doctors to view important images and render important clinical decisions to enhance patient management in an emergency situation. The most important goal of this approach is to minimize the symptom-to-treatment time. Few studies have been published in this regard in the international literature. MMS technology may be an optimum choice for ECG interpretation outside the hospital.

This study was designed to investigate the diagnostic accuracy of interpretations of ECG images taken by mobile phone and sent as a multimedia message to another identical mobile phone.

Materials and Methods

ECG Images

ECGs were taken for 305 consecutive patients presenting to the Başkent University Hospital ED (Adana, Turkey) between November 2008 and January 2009 with complaints including chest pain, shortness of breath, palpitations, or syncope. ECGs were taken by technicians or nurses and were sent to an attending emergency medicine specialist in the ED. Patients' full name and file number were written on the paper ECG traces. Because cardiologists and emergency medicine specialists work 5 days a week, between 8 a.m. and 5 p.m., ECGs taken outside of this time period were not included in the study. The study protocol was approved by the local ethics committee, and all the patients gave written informed consent.

GSM Operator and Mobile Phones

The service of a national operator (Turkcell, Istanbul, Turkey) with more than 84% geographical coverage and two Nokia (Espoo, Finland) N93 mobile phones were used in this study. This operator provided 2.5G (900/1,800 MHz, EDGE) technology service. As of December 31, 2008 the operator has covered 100% of settlements with a population of 1,000 or more.

The mobile phone we chose has advanced camera features. Some data specifications of the mobile phone are shown in Table 1. This camera phone provided four different resolutions. In order to obtain the images of best quality, the highest-resolution setting (2,048×1,536 pixels) was used. Flash, brightness, white balance, and color tone were automatically set by the mobile phone. No external flash unit was used to acquire the images.

Table 1.

Technical Data for the Nokia N93 Mobile Phone

FACTOR	SPECIFICATIONS
Camera	3.15 megapixel, autofocus
Dimensions	118×55.5×28.2 mm
Weight	180 g
Display type	Thin film transistor, 256K colors liquid crystal display
Display size	240×320 pixels, 36×48 mm, 2.4 inches
Operating system	Symbian OS 9.1, S60 3rd edition
Network	UMTS/GSM 900/GSM 1800/GSM 1900

Sending and Displaying of the ECG Images

ECG photographs were taken using the mobile phone camera in JPEG format by the emergency physician. Because the maximum dimension for a multimedia message allowed by the GSM operator was 500 kilobytes, the images were automatically resized by the mobile phone before sending. After resizing, the images were 640×480 pixels. Thereafter, the image data were attached to a multimedia message. The ECG images from the ED were transmitted via MMS to the other identical mobile phone carried by a cardiologist for interpretation. To avoid any negative effect on the functions of medical equipment in the ED, transmissions were performed outside the emergency room.

The ECG images were displayed on the telephone's thin film transistor liquid crystal display and evaluated by the cardiologist (Fig. 1). The zoom function was used as needed for viewing specific leads on the mobile phone display. All ECG images received were classified by the cardiologist according to ability to use them in clinical practice (good, moderate, and poor quality). ECG transmission success was defined as a 12-lead ECG image received on the cardiologist's mobile phone. All ECGs sent and received were stored on both mobile phones for later data documentation.

Fig. 1.

The image of an electrocardiogram sent by multimedia message on a mobile phone screen.

Interpretation of ECG Images

The cardiologist carrying the mobile phone recorded the findings by interpreting the ECG images that were sent to his mobile phone at the end of each day. The paper ECGs were interpreted individually by a staff emergency medicine specialist and another staff cardiologist. The physicians interpreting the ECGs did not have any information regarding the clinical condition of the patients and were unaware of each other's interpretation. The ultimate responsibility belonged to the emergency medicine specialist assessing the patient in the ED. A cardiologist who did not take part in the study was consulted about the patients. Time of delivery of MMS was calculated, taking into account the data of the messages. ECG interpretations were recorded using a programmed response data collection sheet. The duration of ECG interpretation was recorded.

Comparison of the Interpretations

The consensus interpretation of the two cardiologists in the core laboratory was used as the reference ECG interpretation. Misinterpretations included the failure to detect an abnormality (false negative) and the reporting of an abnormality when none existed (false positive). The accuracy of their interpretations was scored on a scale from 1 to 4 using the grading system shown in Table 2.⁶ In the situation where more than one discrepancy was identified, the most significant error was accepted.

Table 2.

Classification of Grades of Agreement

GRADE	DEFINITION	CRITERIA
1	Complete concordance	Accurate interpretation of the electrocardiography
2	Minor disagreement	First-degree heart block, ventricular ectopic beats, sinus tachycardia, sinus bradycardia, minor nonspecific T-wave changes
3	Disagreement	Left ventricular hypertrophy, right or left axis deviation, bundle branch block, slow atrial fibrillation, isolated T-wave inversion or nonspecific ST-segment depression, identification of Q-wave in a lead in which it is not significant, Wolff–Parkinson–White syndrome
4	Significant disagreement, potentially affecting immediate management	ST-segment depression or elevation, widespread T-wave inversion or presence of pathological Q-waves, fast atrial fibrillation or other dysrhythmia

Statistical Analysis

Statistical analysis was performed using a commercially available software package (SPSS version 9.0 for Windows, SPSS, Inc., Chicago, IL). The variables are presented as mean±standard deviation for continuous data and as proportions for categorical data. The levels of agreement between interpreters and the core laboratory were evaluated using the κ statistic. This test shows the level of agreement between observers that is better than chance alone. Thus, perfect agreement is 1.0, and agreement by chance alone is 0. Table 3 illustrates a generally accepted mode of κ value interpretation. The distribution of error gradings was compared by a chi-squared McNemar test. The duration of interpretation was compared by an independent-samples t-test. Two-sided p values of <0.05 were considered significant.

Table 3.

Generally Accepted Mode of κ Value Interpretation

VALUE OF κ	STRENGTH OF AGREEMENT
≤0.20	Poor
0.21–0.40	Fair
0.41–0.60	Moderate
0.61–0.80	Good
0.81–1.00	Very good

Results

Image and Transfer Information

In the study 305 ECGs were used. All ECGs were successfully transmitted on the first attempt.

The success rate for receiving multimedia messages was 100%. The mean transmission time was 1.08±0.27 min. Of the total number of ECGs sent, 277 (90.8%) were of good quality, 28 (9.2%) were of moderate quality, and none (0%) was of poor quality (i.e., impossible or almost impossible to analyze). The mean size of ECG images before adding a multimedia message was 970.95±115.32 kilobytes. The mean size of ECG images sent by MMS after zooming out was 44.90±7.45 kilobytes.

Duration of Interpretation

The mean duration of ECG interpretation of the core laboratory was 9.95±0.73 s, for the cardiologist interpreting the paper ECGs 10.06±0.78 s, for the emergency medicine specialist 10.04±0.51 s, and for the cardiologist interpreting on the mobile phone screen 59.70±5.69 s. Compared with the core laboratory, there was no significant difference between the cardiologist and emergency medicine specialist interpreting the paper ECG (p=0.082 and p=0.092, respectively), but interpretations made on a mobile phone screen took significantly longer (p<0.001).

Comparison of ECG Interpretations

The core laboratory reported that of the 305 ECGs, 76 (24.9%) were completely normal. Compared with the findings of the core laboratory according to the scale defined in advance, error grades of the interpretations concerning overall concordance are shown in Table 4. The cardiologist interpreting the paper ECG made significantly fewer Grade 4 errors than both the emergency medicine specialist and the cardiologist interpreting the ECG on the phone screen (p=0.023 and p=0.025, respectively). There was no statistically significant difference observed between the emergency medicine specialist and the cardiologist interpreting the ECG on the phone screen with respect to Grade 4 error (p=0.76).

Table 4.

Error Grades of Interpreters According to the Core Laboratory Findings

	N (%)
ERROR GRADE	CARDIOLOGIST INTERPRETING THE PAPER ECG	CARDIOLOGIST INTERPRETING ECG ON THE PHONE SCREEN	EMERGENCY MEDICINE SPECIALIST INTERPRETING THE PAPER ECG
1	216 (70.8)	169 (55.4)	143 (46.9)
2	43 (14.1)	52 (17.0)	56 (18.4)
3	32 (10.5)	58 (19.0)	77 (25.2)
4	14 (4.6)	26 (8.5)^a,b	29 (9.5)^c

p=0.023, Grade 4 versus Grade 1, 2, and 3 errors versus the cardiologist interpreting the paper electrocardiogram (ECG).

p=0.76, Grade 4 versus Grade 1, 2, and 3 errors versus the emergency medicine specialist.

p=0.025, Grade 4 versus Grade 1, 2, and 3 errors versus the cardiologist interpreting the paper ECG.

Interobserver κ coefficients for ECG findings and positive and negative predictive values are shown in Table 5. Considering the degree of agreement in grading the entirely normal ECGs, it was observed that the cardiologist interpreting the paper ECG and the cardiologist interpreting the phone screen showed a significant agreement (κ=0.80 and κ=0.69, respectively), whereas the emergency medicine specialist showed a moderate agreement (κ=0.44).

Table 5.

Interobserver Agreement and Positive and Negative Predictive Values of Electrocardiogram Interpretation

		CARDIOLOGIST INTERPRETING THE PAPER ECG			CARDIOLOGIST INTERPRETING ECG ON THE PHONE SCREEN			EMERGENCY MEDICINE SPECIALIST INTERPRETING THE PAPER ECG
ECG FINDINGS	F (N)	κ	PPV (%)	NPV (%)	κ	PPV (%)	NPV (%)	κ	PPV (%)	NPV (%)
Sinus rhythm	260	1.0	100	100	0.95	99.2	95.6	0.90	99.2	87.8
Entirely normal ECG	76	0.80	81.9	96.3	0.69	70.6	94.8	0.44	64.4	85.6
Minor T-wave changes	49	0.38	48.9	89.9	0.23	47.8	86.5	0.15	30.0	86.0
Sinus tachycardia	46	0.86	81.5	99.2	0.86	88.9	97.7	0.84	82.3	98.4
Pathological Q-waves^a	39	0.93	94.7	98.8	0.85	89.2	97.8	0.87	89.5	98.1
Isolated T-wave inversion or nonspecific ST-segment depression	37	0.60	90.0	93.3	0.43	57.1	92.4	0.42	48.6	92.9
Atrial fibrillation	35	1.0	100	100	0.94	94.3	99.3	0.89	85.0	99.6
RBBB	24	0.91	91.7	99.3	0.83	90.5	98.2	0.84	84.0	98.9
ST-segment depression^a	24	0.91	91.7	99.2	0.81	86.3	98.2	0.77	81.8	97.8
LAFB	21	0.71	64.2	98.9	0.70	77.8	97.6	0.61	51.4	98.9
Q-wave in a lead	21	0.86	100	98.3	0.52	50.0	97.1	0.58	56.0	97.5
Fast atrial fibrillation^a	19	0.97	94.7	100	0.91	94.4	99.3	0.91	94.4	99.3
Left axis deviation	19	0.80	83.3	98.6	0.64	60.9	98.2	0.49	39.0	98.8
ST-segment elevation^a	18	0.91	94.1	99.3	0.81	88.8	98.6	0.73	76.4	98.2
Slow atrial fibrillation	16	0.97	100	99.7	0.78	82.4	98.6	0.86	76.2	100
Ventricular ectopic beat	15	1.0	100	100	1.0	100	100	0.93	93.3	99.7
LBBB	11	1.0	100	100	0.91	90.9	99.7	0.60	45.4	99.6
Sinus bradycardia	8	0.87	87.5	99.7	0.82	77.8	99.7	0.87	87.5	99.7
Widespread T-wave inversion^a	8	0.94	88.8	100	0.87	87.5	99.7	0.85	100	99.3
SVT^a	8	1.0	100	100	1.0	100	100	0.80	85.7	99.3
Left ventricular hypertrophy	8	0.87	80.0	100	0.52	38.9	99.7	0.53	42.8	99.3

The findings lead to Grade 4 error.

F, the frequency of ECG findings at the core laboratory; LAFB, left anterior fascicular block; LBBB, left bundle branch block; NPV, negative predictive value; PPV, positive predictive value; RBBB, right bundle branch block; SVT, supraventricular tachycardia.

Considering major ECG findings, when misinterpretation of the ECG was considered as Grade 4 error, the degree of agreement of the emergency medicine specialist was good (κ=0.72, κ=0.77, and κ=0.80, respectively) and that of the cardiologists interpreting paper ECG and MMS on the phone screen was superior (κ>0.81) in detecting ST-segment elevation, ST-segment depression, and supraventricular tachycardia. The degree of agreement between the three specialists overall was strong (κ>0.81) in detecting pathological Q-waves, widespread T-wave inversion, and fast atrial fibrillation. ECG findings showing the least agreement for all three specialists was minor T-wave changes.

There were 18 ECGs reported to have ST-segment elevation by the core laboratory. In terms of determining ST-segment elevation the number of false-negative ECGs for the cardiologist interpreting paper ECG was two, for the cardiologist interpreting on the phone screen was four, and for the emergency medicine specialist was five. The number of false-positive ECGs was one, two, and four, respectively. The positive predictive values in determining ST-segment elevation were 94.1%, 88.8%, and 76.4%, respectively.

Discussion

It has been shown that wireless mobile systems are useful and reliable for the transfer of ECG images taken outside the hospital to hospitals and from there to handheld computers or mobile phones of cardiac consultant specialists. However, the cost of these systems is high, and this limits their common use. More than a third of ambulances have the equipment to be able to transfer prehospital ECGs to hospitals in Europe and in the United States.¹³ Using the method in our study, expensive equipment is not needed. The total cost of the two mobile phones used in this study was about $2,000, and that of each MMS was 50 cents (U.S. $1=TL 1.60).

There are a limited number of studies about using MMS to send ECG images. In a small study,¹² ECG images of 10 patients with acute coronary syndrome acquired with the camera of a mobile phone were sent using MMS to an e-mail account, and the cardiologist interpreted the ECGs via cell phone. In the aforementioned study the images were sent approximately in 71.5 s, and the cardiologist interpreted them in around 24 s. The results obtained showed that a more rapid diagnosis was established than by the fax device for patients with acute coronary syndrome. Other ECG findings were not assessed.¹² Such a practice requires the use of Internet access and an e-mail account. However, in our study MMS and the GSM signal were used. Although mean transfer time (1.08±0.27 min) was similar, interpretation on the cell phone screen was determined to be longer (59.70±5.69 s). The reason for this might be that it was necessary for us to record not only ST-segment changes but also all ECG findings in our study. Given that GSM signals are more widespread than Internet connections, the use of MMS can be considered to be more advantageous for sending ECGs remotely.

Although there was no significant difference in terms of total grade 4 error rate between the cardiologist making interpretations on the mobile phone screen and the emergency medicine specialist, it can be said that teleconsultation may provide, albeit a small, additional benefit in determining ST-segment changes and supraventricular tachycardias. But, the use of this method for the transfer of the ECG together with the patient's medical history may be considered to provide more benefit for the purpose of teleconsultation in healthcare facilities where there is neither a cardiologist nor an emergency medicine specialist or in ambulances with no ECG transfer equipment.

It was shown that transmission of prehospital ECGs directly to the attending cardiologist's mobile phone decreased door to percutaneous coronary intervention time in ST-segment elevation myocardial infarction patients.¹⁴ Given transfer success and the duration of the transfer and interpretation, being able to consult an ECG within minutes via MMS in any location where the GSM signal reaches may be beneficial in referring patients experiencing ST-segment elevation myocardial infarction for primary percutaneous coronary intervention and for performing expeditious treatment. In addition, this method is also very practical for ECG consultations outside the emergency setting. With the combining of wireless data transmission systems, together with an advanced mobile operating system, portable mobile devices with a camera and screen features may allow more reliable, inexpensive, and simple teleconsultation practice in future years.

Limitation of the Study

The suitability of ECG images received by mobile phone for analysis was evaluated to be subjective by the cardiologist who made the interpretations.

Conclusions

Transmission of ECG images acquired via the camera of a mobile phone as a MMS and interpretation of these images on the mobile phone screen is an affordable and simple practice of telecardiology and can be used as an aide in ECG consultation.

Footnotes

Disclosure Statement

No competing financial interests exist.

References

Cannon

, Gibson

, Lambrew

et al. Relationship of symptom-onset-to-balloon time and door-to-balloon time with mortality in patients undergoing angioplasty for acute myocardial infarction. JAMA, 2000; 283:2941–2947.

Salerno

, Alguire

, Waxman

. Training and competency evaluation for interpretation of 12-lead electrocardiograms: Recommendations from the American College of Physicians. Ann Intern Med, 2003; 138:747–750.

Morrison

, Verbeek

, McDonald

, Sawadsky

, Cook

. Mortality and prehospital thrombolysis for acute myocardial infarction: A meta-analysis. JAMA, 2000; 283:2686–2692.

Lee

, Weisberg

, Brand

, Rouan

, Goldman

. Candidates for thrombolysis among emergency room patients with acute chest pain. Potential true- and false-positive rates. Ann Intern Med, 1989; 110:957–962.

Snoey

, Housset

, Guyon

, ElHaddad

, Valty

, Hericord

. Analysis of emergency department interpretation of electrocardiograms. J Accid Emerg Med, 1994; 11:149–153.

White

, Woodmansey

, Ferguson

, Channer

. Improving the interpretation of electrocardiographs in an accident and emergency department. Postgrad Med J, 1995; 71:132–135.

Grim

, Feldman

, Martin

, Donovan

, Nevins

, Childers

. Cellular telephone transmission of 12-lead electrocardiograms from ambulance to hospital. Am J Cardiol, 1987; 60:715–720.

Pettis

, Savona

, Leibrandt

et al. Evaluation of the efficacy of hand-held computer screens for cardiologists' interpretations of 12-lead electrocardiograms. Am Heart J, 1999; 138:765–770.

Davis

, Graydon

, Stein

et al. The positive predictive value of paramedic versus emergency physician interpretation of the prehospital 12-lead electrocardiogram. Prehosp Emerg Care, 2007; 11:399–402.

10.

Leibrandt

, Bell

, Savona

et al. Validation of cardiologists' decisions to initiate reperfusion therapy for acute myocardial infarction with electrocardiograms viewed on liquid crystal displays of cellular telephones. Am Heart J, 2000; 140:747–752.

11.

Campbell

, Patterson

, Cromer

et al. Prehospital triage of acute myocardial infarction: wireless transmission of electrocardiograms to the on-call cardiologist via a handheld computer. J Electrocardiol, 2005; 38:300–309.

12.

Ohtsuka

, Uchida

, Nakajima

, Yamaguchi

, Takano

, Komuro

. Transferring images via the wireless messaging network using camera phones shortens the time required to diagnose acute coronary syndrome. Circ J, 2007; 71:1499–1500.

13.

Bassand

, Danchin

, Filippatos

et al. Implementation of reperfusion therapy in acute myocardial infarction. A policy statement from the European Society of Cardiology. Eur Heart J, 2005; 26:2733–2741.

14.

Sejersten

, Sillesen

, Hansen

et al. Effect on treatment delay of prehospital teletransmission of 12-lead electrocardiogram to a cardiologist for immediate triage and direct referral of patients with ST-segment elevation acute myocardial infarction to primary percutaneous coronary intervention. Am J Cardiol, 2008; 101:941–946.