Automatic ECG quality scoring methodology: mimicking human annotators

An algorithm to determine the quality of electrocardiograms (ECGs) can enable inexperienced nurses and paramedics to record ECGs of sufficient diagnostic quality. Previously, we proposed an algorithm for determining if ECG recordings are of acceptable quality, which was entered in the PhysioNet Challenge 2011. In the present work, we propose an improved two-step algorithm, which first rejects ECGs with macroscopic errors (signal absent, large voltage shifts or saturation) and subsequently quantifies the noise (baseline, powerline or muscular noise) on a continuous scale. The performance of the improved algorithm was evaluated using the PhysioNet Challenge database (1500 ECGs rated by humans for signal quality). We achieved a classification accuracy of 92.3% on the training set and 90.0% on the test set. The improved algorithm is capable of detecting ECGs with macroscopic errors and giving the user a score of the overall quality. This allows the user to assess the degree of noise and decide if it is acceptable depending on the purpose of the recording.     

Does Atrial Fibrillation Cause False-Positive Late Potentials?

We hypothesized that atrial fibrillation may cause false-positive late potentials due to the recording of baseline atrial activity. We performed signal-averaged ECGs in 26 patients with atrial fibrillation before and after conversion to normaJ sinus rhythm. Signal-averaged ECGs were recorded for > 200 cycles with a noise level of ≤ 0.5 μV. The signals were band-pass filtered at 40–250 Hz. We examined filtered QRS duration (fQRS), duration of low amplitude signal < 40 μV (LAS), and the root mean square (RMS) of the terminal 40 msec of the QRS complex. A late potential was considered present when two of the following three criteria were met: fQRS ± 114 msec, LAS ± 38 msec, and RMS ≤ 20 μV. The mean ± standard deviation of the fQRS in atriai fibrillation and sinus rhythm were 313 ± 28 and 110 ± 25 msec; of the LAS 38 ± 17 and 37 ± 15 msec; of the RMS 27 ± 22 and 28 ± 21 μV; of the noise 0.25 ± 0.08 and 0.22 ± 0.07 μV (P = NS). Ten signal-averaged ECGs in atrial fibrillation had late potentials. With reversion to sinus rhythm one of these 26 patients gained a late potential; two others lost a late potential (P = NS by M c Nemar's Chi-square). There was no significant difference in the signal-averaged ECG parameters or noise levels. In conclusion, signal-averaged ECG parameters are not significantly changed by cardioversion of atrial fibrillation to normal sinus rhythm. Moreover, the presence of a late potential in a patient with atrial fibrillation does not appear to be the result of a spurious recording of baseline atrial fibrillatory activity.     

Signal quality indices and data fusion for determining clinical acceptability of electrocardiograms

A completely automated algorithm to detect poor-quality electrocardiograms (ECGs) is described. The algorithm is based on both novel and previously published signal quality metrics, originally designed for intensive care monitoring. The algorithms have been adapted for use on short (5-10?s) single- and multi-lead ECGs. The metrics quantify spectral energy distribution, higher order moments and inter-channel and inter-algorithm agreement. Seven metrics were calculated for each channel (84?features in all) and presented to either a multi-layer perceptron artificial neural network or a support vector machine (SVM) for training on a multiple-annotator labelled and adjudicated training dataset. A single-lead version of the algorithm was also developed in a similar manner. Data were drawn from the PhysioNet Challenge 2011 dataset where binary labels were available, on 1500 12-lead ECGs indicating whether the entire recording was acceptable or unacceptable for clinical interpretation. We re-annotated all the leads in both the training set (1000 labelled ECGs) and test dataset (500 12-lead ECGs where labels were not publicly available) using two independent annotators, and a third for adjudication of differences. We found that low-quality data accounted for only 16% of the ECG leads. To balance the classes (between high and low quality), we created extra noisy data samples by adding noise from PhysioNet's noise stress test database to some of the clean 12-lead ECGs. No data were shared between training and test sets. A classification accuracy of 98% on the training data and 97% on the test data were achieved. Upon inspection, incorrectly classified data were found to be borderline cases which could be classified either way. If these cases were more consistently labelled, we expect our approach to achieve an accuracy closer to 100%.     

Neural networks--a diagnostic tool in acute myocardial infarction with concomitant left bundle branch block

The prognosis of acute myocardial infarction (AMI) improves by early revascularization. However the presence of left bundle branch block (LBBB) in the electrocardiogram (ECG) increases the difficulty in recognizing an AMI and different ECG criteria for the diagnosis of AMI have proved to be of limited value. The purpose of this study was to detect AMI in ECGs with LBBB using artificial neural networks and to compare the performance of the networks to that of six sets of conventional ECG criteria and two experienced cardiologists. A total of 518 ECGs, recorded at an emergency department, with a QRS duration > 120 ms and an LBBB configuration, were selected from the clinical ECG database. Of this sample 120 ECGs were recorded on patients with AMI, the remaining 398 ECGs being used as a control group. Artificial neural networks of feed-forward type were trained to classify the ECGs as AMI or not AMI. The neural network showed higher sensitivities than both the cardiologists and the criteria when compared at the same levels of specificity. The sensitivity of the neural network was 12% (P = 0.02) and 19% (P = 0.001) higher than that of the cardiologists. Artificial neural networks can be trained to detect AMI in ECGs with concomitant LBBB more effectively than conventional ECG criteria or experienced cardiologists.     

Feasibility of chronic pulmonary heart diagnosis by ECG criteria

AIM: To determine diagnostic value of the method of ECG in 12 conventional leads for diagnosis of pulmonary heart. MATERIALS AND METHODS: The ECG data for 41 patients with clinical diagnosis "pulmonary heart", a risk group of 59 patients with suspected diagnosis "pulmonary heart" and healthy residents of the Komi Republic (n = 968) and Arkhangelsk region (n = 62) have been analysed. The ECGs were recorded using standard electrocardiographs with manual measurement of the amplitude and temporal parameters. RESULTS: We have analysed 215 diagnostic ECG criteria of "pulmonary heart" used in practice, selected 19 informative indices and developed 19 new ones. On the basis of 17 most significant criteria a short algorithm and a program of diagnosis of "pulmonary heart" for IBM PC have been elaborated. The test of the algorithm efficiency by a blind method confirmed the clinical diagnosis in 81.3% of cases, hypodiagnosis and hyperdiagnosis in 6.2 and 12.5%, respectively. CONCLUSION: A widespread and noninvasive ECG method proved promising in diagnosis of pulmonary heart.     

Prevalence and clinical course of the juveniles with Brugada-type ECG in Japanese population

BACKGROUND: Although many studies on Brugada syndrome have been done, with many reports of genetic findings and clinical features, little evidence exists to support the role of this syndrome in sudden cardiac death in a juvenile population. We sought to determine the prevalence and clinical course in children exhibiting Brugada-type ECG in a community-based population. METHODS: Our study population comprised 21,944 subjects (11,282 boys and 10,662 girls) who underwent ECG during their first-year elementary school health examinations between 1992 and 2001 in Izumi City, Osaka. Brugada-type ECG was defined as demonstrating ST-segment elevation (coved or saddle-back type, J wave amplitude > or =0.2 mV) in the right precordial leads. We also divided Brugada-type ECGs into three types according to a consensus report. Type 1: coved ST-segment elevation displaying high J wave amplitude followed by a negative T wave; Type 2: high take-off and gradually descending ST-segment elevation (remaining > or =1 mm) followed by a positive or biphasic T wave; and Type 3: ST-segment elevation of <1 mm of both types. RESULTS: Four subjects showed Brugada-type ECG (0.02%) (2 boys and 2 girls). Only one subject, a girl, met Type 1 criteria (0.005%). No history of structural heart disease was documented in these four subjects. During 6.8 +/- 1.0 years of follow-up, no episode of unexpected sudden death, syncopal attack, and fatal arrhythmia occurred. CONCLUSIONS: The prevalence of Brugada-type ECG in a juvenile population was extremely low. To investigate when the typical Brugada-type ECG might be manifested, it could be necessary to check ECGs after adolescence.     

Determination of the ability of high-frequency ECG to estimate left ventricular mass in humans, determined by magnetic resonance imaging

BACKGROUND: Previous studies have shown a significantly higher correlation between left ventricular mass index (LVMi) and high-frequency QRS components (HF-QRS) than between LVMi and QRS amplitudes in the standard frequency range in rabbits. The purpose of the present study was to compare ECG measurements from standard and high-frequency ranges with left ventricular mass (LVM) and LVMi determined by magnetic resonance imaging in humans. METHODS: Sixty-two normal subjects were studied. Signal-averaged ECGs from the 12 standard leads were analysed in the standard frequency range (0.05-150 Hz), in the middle (25-100 Hz) and high end (50-150 Hz) of the standard frequency range and in the 150-250 Hz range. Root-mean square (RMS) values from the HF-QRS and QRS amplitude measurements from the standard ECGs were compared with LVM and LVMi. RESULTS: The correlations between LVMi and HF-QRS were similar to those between LVMi and standard ECG. When regarding LVM, however, the correlations found in the standard ECG were higher than those found in HF-QRS. CONCLUSIONS: Contrary to previous results in animals, we found in humans no better correlation between HF-QRS and LVM/LVMi than between standard ECG and LVM/LVMi.     

Serial Electrocardiograms for Chest Pain Patients with Initial Nondiagnostic Electrocardiograms: Implications for Thrombolytic Therapy

Objectives: To determine the proportion of acute myocardial infarction (AMI) patients without ST–segment elevation who subsequently develop ST–segment elevation during their hospital courses; and to compare demographics and presenting features of AMI patient subgroups: those with initial ST–segment elevation, those with in–hospital ST–segment elevation, and those with no ST–segment elevation. Methods: A retrospective cohort analysis of admitted chest pain patients who had a hospital discharge diagnosis of AM1 was performed. Each chart was examined for initial ECG interpretation, serial ECG analysis, patient age, gender, cardiac risk factors, in-hospital survival, time between sequential ECGs, and number of ECGs performed within the first 48 hours of hospital admission. Results: Of the 114 charts reviewed, 20 patients had ECGs meeting thrombolytic criteria on arrival. Of the 94 AM1 patients who had nondiagnostic ECGs on arrival, 19 (20%) subsequently developed ECG changes meeting thrombolytic criteria. Seven patients developed these changes within eight hours of the initial ECG, four from eight to 12 hours after, two from 12 to 24 hours after, and six more than 24 hours after. Most patients who had documented AMIs did not develop ECG criteria for thrombolytic therapy during their hospitalizations. Male gender and smoking history were more commonly associated with late ST-segment elevation for those presenting with nondiagnostic ECGs. All the patients who had late diagnostic ECG changes survived to hospital discharge. Serial ECGs were performed more frequently in the group who had initially diagnostic ECGs and least frequently in the group who did not develop ST-segment elevation during their hospitalizations. Conclusions: Most patients with AM1 do not meet ECG criteria for the administration of thrombolytic therapy. A significant minority (20%) of the admitted chest pain patients with subsequently confirmed AMIs developed ECG criteria for thrombolytics during their hospitalizations. Further attention to such patients who have delayed ST-segment elevation is warranted. A standardized in-hospital serial ECG protocol should be considered to identify admitted patients who develop criteria for thrombolytic or other coronary revascularization therapy.     

Assessment of Noise in Digital Electrocardiograms

BATCHVAROV, V., et al. : Assessment of Noise in Digital Electrocardiograms. Technically related noise in 12-lead ECGs recorded with ambulatory recorders has never been systematically compared with that in ECGs recorded with conventional ECGs. This study compared serial 10-second ECGs obtained in ten healthy men, age 22–45 years, who were recorded in the supine resting position using a (1) MAC VU recorder, (2) digital ambulatory SEER MC recorder with a Multi-Link detachable ECG cable, and (3) digital ambulatory SEER MC recorder with a light ambulatory ECG cable. In each ECG, averaged sinus rhythm cycles of the entire recording were realigned with the native signal and subtracted. The resulting "residuum" was quantified by computing its standard deviation and root mean square of successive differences (RMSSD). While the RMSSD residuum values were significantly lower with the MAC VU recorder (  6.27 ± 0.98 μV  ) than with the SEER MC recorder with either ECG cables (  7.29 ± 1.31 and 7.17 ± 1.31 μV, P < 0.003 and p < 0.02  ), the difference was practically negligible and there was no detectable difference in the standard deviation residuum values. The study concludes that valid ECG investigations of serial ECG testing may be conducted using the ambulatory SEER MC recorders providing the biological sources of ECG noise are controlled. The available technology for noise assessment suggests that studies involving advanced analysis of serial ECGs (e.g., of drug related changes), should incorporate objective characterisation of ECG quality.     

Holter Recordings with Continuous Marker Annotations: A New Tool in Pacemaker Diagnostics

Pacemakers provide marker annotations to facilitate the interpretation of pacemaker electrocardiograms (ECGs) and can be used in cases of suspected pacemaker malfunction or to understand pacemaker behavior. Due to the need for a programmer, only short-term evaluations are possible. We evaluated a prototype Telemetry Data Logger (TDL) designed to continuously transfer markers from the pacemaker to a conventional Holter recorder. A miniaturized telemetry receiving coil was attached to patient's skin above the pacemaker, which was programmed to transmit markers continuously. The TDL, which receives and converts markers into eight positive and eight negative deflections, ranging from -2.5 to +2.5 mV in amplitude, was connected to one channel of a conventional Holter recorder (Tracker 2). We performed 20 Holters in 13 patients who had implanted VDDB or DDDR devices from the same manufacturer and evaluated three versions of software. Marker transmission was possible in all patients, producing Holter ECGs with complete marker annotations. Artifacts occurred < 4 % of the time. A 50-ms rectangular pulse was optimal for marker interpretation. The device, which was easy to use and well accepted by the patients, assisted in the diagnosis of inappropriate pacemaker programming, even when the surface ECG seemed to show regular pacemaker function. In the presence of low quality surface ECGs, marker annotations allowed the assessment of pacemaker function. The capability to annotate the onset of special algorithms, like tachycardia termination algorithms or mode switching, facilitates interpretation of pacemaker behavior, enabling a reliable assessment of the appropriateness of such algorithms. Conclusion: The TDL effectively enables pacemaker markers to be inscribed onto a conventional Holter recording, facilitating the interpretation of pacemaker ECGs and the diagnosis of inappropriate pacemaker programming even when not discernible from the surface ECG alone.     

T/QRS ratio best distinguishes ventricular aneurysm from anterior myocardial infarction

OBJECTIVES: Reperfusion therapy for acute myocardial infarction (AMI) is indicated in the presence of ST elevation (STE) and ischemic symptoms. Previous MI may present with persistent STE or "left ventricular aneurysm" (LVA) morphology that mimics AMI. Hypothesis A high ratio of T amplitude to QRS amplitude best distinguishes AMI from LVA. METHODS: This was a retrospective cohort analysis. Patients with anatomical LVA by echocardiography were identified and those who presented to the ED with ischemic symptoms and STE of at least 1 mm in 2 consecutive leads and ruled out for acute left anterior descending coronary artery (LAD) occlusion were selected. Electrocardiograms (ECGs) were compared with a control group of 37 consecutive anterior AMI (aAMI) with proven acute LAD occlusion. Bundle-branch block was excluded. Various ECG measurements and ratios were compared. RESULTS: Twenty patients with LVA met the inclusion criteria. The best discriminator was T amplitude sum to QRS amplitude sum ratio V1-V4, misclassifying only 4 (6.8%) of 59 cases at a cutoff of >0.22 for AMI. For aAMI and LVA, respectively, mean (+/-95% CI) ratio of the sum of T amplitudes in V 1 to V 4 to the sum of QRS amplitude in V1-V4 was 0.54+/-0.085 and 0.16+/-0.021 (P<.00012). Thirty-five of 37 aAMI had a ratio>0.22; the false negatives (ratio<0.22) had 11.5 and 6 hours of symptoms before the ECG. Twenty of 22 LVA had a ratio相似文献   

The Comparative Predictive Accuracy of Spectral Turbulence Analysis and Time-Domain Analysis for Sustained Monomorphic Ventricular Tachycardia in Postinfarction Patients with Conduction Defects

Signal-averaged electrocardiograms obtained in 86 postinfarction patients with right bundle branch block (RBBB), left bundle branch block (LBBB), or intraventricular conduction defect (IVC-D), underwent time-domain analysis (TDA) and spectral turbulence analysis (STA) to determine which approach provided the more effective marker for patients with sustained monomorphic ventricular tachycardia. TDA parameters included the root mean square value of the last 40 ms of the vectormagnitude complex and the duration of the low amplitude signal below 40 μV STA utilized a summation lead (X + Y + Z) and quantitated four parameters: interslice correlation mean, interslice correlation standard deviation, low slice correlation ratio, and spectral entropy. High-pass filters of 40 Hz and 25 Hz were used to study the total patient population with noise levels ≤1 μV and a subset of 67 patients with noise levels ≤ 0.5 μV, The techniques compared their effectiveness as measured by their positive predictive values (PPV), negative predictive values (NPV), sensitivity (Sn), and specificity (Sp). In RBBB, STA was uniformly a more powerful tool utilizing either filter at both noise levels. In LBBB, STA was consistently more powerful at both noise levels at 40 Hz and, generally, more powerful at 25 Hz with isolated exceptions. In conduction defects in which QRS was > 100 ms but < 120 ms, TDA was equal to or more effective than STA, with the exception of PPV and Sp at 40 Hz at 1-μV noise level and the Sp at 0.5 μV. The addition of ejection fraction data to STA score resulted in further overall improvement in performance, but above conclusions were unchanged.     

Is chronic kidney disease comparable to diabetes as a coronary artery disease risk factor?

BACKGROUND: Chronic kidney disease (CKD) is one of the known risk factors for coronary heart disease (CHD). Though electrocardiograms (ECGs) have limited accuracy in determining the true prevalence of CHD, we wondered whether CKD and diabetes mellitus (DM) controlled for hypertension (HTN), had similar prevalences of ECG abnormalities that could reflect underlying coronary heart disease. METHOD: Data were collected for 5,942 men and women aged 30 to 69 years in the Tehran Lipid and Glucose Study (TLGS), a crosssectional phase of a large epidemiologic study first initiated in 1999. ECG findings of all subjects were coded according to Minnesota ECG coding criteria. The Whitehall criteria for abnormal ECG findings that could represent ischemia were utilized. Creatinine clearance (Crcl) was estimated using the Cockroft-Gault equation and diabetes was defined according to the American Diabetic Association (ADA) criteria. Subjects with moderate CKD and without DM were compared with the patients with DM without CKD. HTN prevalence was similar. The analysis was performed for all Whitehall ECG ischemia abnormalities combined, and separately for pathologic Q waves. RESULTS: In spite of an overall similar prevalence of smoking, and a lower incidence of dyslipidemia and HTN, moderate CKD patients had a higher prevalence of Whitehall criteria abnormal ECG findings compared with the patients with DM. Over 19% of patients with CKD had abnormal ECG findings while 14.7% of diabetic patients had abnormal ECGs (P = 0.02). The prevalence of Q waves was 11.5% in patients with CKD and 10.8% in patients with DM. In an age-matched subgroup of patients with DM and no CKD, the prevalence of ECG abnormalities was 19.3%, similar to the patients with moderate CKD and no DM (19.7%) (P = 0.9). The prevalence of pathologic Q waves in an age-matched group was 11.45%, compared with 11.5%, respectively. CONCLUSION: Moderate CKD is a major risk factor for the development of the Whitehall ECG criteria which have been associated with ischemic heart disease. The importance of CKD as a risk factor for ECG abnormalities is comparable with DM. Patients with moderate CKD probably are candidates for aggressive CHD risk modification.     

ECG Leads I and II to Evaluate Diuresis of Patients with Congestive Heart Failure Admitted to the Hospital via the Emergency Department

Background: Attenuation of electrocardiogram (ECG) QRS complexes is observed in patients with a variety of illnesses and peripheral edema (PERED), and augmentation with alleviation of PERED. Serial ECGs in stable individuals display variation in the amplitude of QRS complexes in leads V1–V6, stemming from careless placement of recording electrodes on the chestwall. Electrocardiographs record only leads I and II, and mathematically derive the other four limb leads in real time. This study evaluated the sum of the amplitudes of ECG leads I and II, along with other sets of ECG leads in the monitoring of diuresis in patients with congestive heart failure (CHF).
Methods: Twenty patients with CHF had ECGs and weights recorded on admission and at discharge. The amplitude of the QRS complexes in all ECG leads were measured and sums of I and II, all limb leads, V1–V6, and all 12 leads were calculated.
Results: There was a good correlation between the weight loss and the increase in the sums of the amplitudes of the QRS complexes from leads I and II (r = 0.55, P = 0.012), and the six limb leads (r = 0.68, P = 0.001), but a poor correlation with the V1–V6 leads (r = 0.04, P = 0.85) and all 12 leads (r = 0.1, P = 0.40).
Conclusions: Sums of the amplitudes of the ECG QRS complexes from leads I and II constitute a reliable, easily obtainable, ubiquitously available, bedside clinical index, which can be employed in the diagnosis, monitoring of management, and follow-up of patients with CHF.     

Electrocardiographic artifacts due to electrode misplacement and their frequency in different clinical settings

Misplacement of electrodes can change the morphology of an electrocardiogram (ECG) in clinical important ways. To assess the frequency of these errors in different clinical settings, we collected ECGs routinely performed at the cardiology outpatient clinic and the intensive care unit. Lead misplacement was suspected when one of the following morphological changes occurred: QRS axis between 180 degrees and -90 degrees , positive P wave in lead aVR, negative P waves in lead I and/or II, very low (<0.1 mV) amplitude in an isolated peripheral lead, or abnormal R progression in the precordial leads. We analyzed 838 ECGs and identified 37 ECGs suspicious for electrode misplacement, from which 7 were confirmed. The frequency of ECG artifacts due to switched electrodes was 0.4% (3/739) at the outpatient clinic and 4.0% (4/99) at the intensive care unit (P = .005). In conclusion, errors in ECG performance do occur with an increasing frequency in an acute medical care setting.     

Analysis of the electrocardiographic waveforms produced by right ventricular pacing: relation to the nonpaced patterns

Background: Ventricular aberrant conduction has a confounding effect on the known relationships between the electrocardiogram (ECG) and left ventricular (LV) mass. By relating the ECG of right ventricular pacing to LV mass and to nonpaced recordings, clarification of these effects might emerge.
Methods and Results: In 30 patients (age, 81 ± 7 years; 13 women) who had right ventricular paced ECGs and echocardiograms, 24 of who also had nonpaced ECGs, comparative analyses were performed. Although the nonpaced ECGs had strong correlations with various echocardiographic measurements, for paced ECGs, only QRS complex voltage and interventricular septal thickness (IVS) were significantly related. However, paced QRS complex voltage relationships correlated with those of nonpaced QRS complexes, ranging from an r = 0.54, P < 0.006, for the sum of the R in aVL and the S in V-3 to r = 0.78, P < 0.001, for the sum of the R in I and the S in III. Paced ECGs had a QRS complex with a greater spatial amplitude, a longer duration, and a more superior position, and had more deeply inverted T waves than nonpaced ECGs. The differences between the voltages of paced and nonpaced QRS complexes, moreover, diminished as LV mass and/or IVS increased. When compared with nonpaced ECGs, paced ECGs showed the most similarity to nonpaced ECGs having a left bundle branch block (LBBB) pattern. Except for the presence of more superiorly directed QRS complexes, paced impulses were not significantly different (P < 0.008) from nonpaced impulses having a LBBB pattern. Also, the nonpaced ECG pattern had no discernable effect on ECG produced by right ventricular (RV) pacing.
Conclusions: Despite having weak relations with echocardiographic measurements, the QRS complex voltage of the paced ECG correlated with those of nonpaced ECGs, and the voltage differences between them were smaller as LV mass increased.     

Diastolic Potentials Recorded by Surface Electrocardiographic Signal Averaging During Sustained Ventricular Tachycardia: Possible Origin From the Reentrant Circuit

ECG signal averaging can detect low amplitude diastolic potentials in sinus rhythm. We, therefore, recorded signal-averaged ECGs during eight episodes of inducible uniform sustained VT with coincident atrial pacing to look for continuous diastolic electrical activity. Simultaneous AV pacing in seven patients served as controls. The number of QRS complexes averaged (187 +/- 47 vs 183 +/- 63), the noise level (1.26 +/- 0.88 vs 1.39 +/- 0.47) and cycle length (385 +/- 52 vs 404 +/- 40) did not differ between VT and paced recordings. In each lead the difference in onset between the unfiltered surface recording and the filtered data (40 Hz bidirectional) was significantly greater in VT than the paced recordings (25 +/- 16 vs 11 +/- 8 msec, P = 0.0012). These late diastolic (pre-QRS) potentials were greater than 15 msec duration in 65% of the leads in VT versus 20% of paced recording (P = 0.021). The maximum value was greater than 20 msec in six VT (75%) versus one (14%) paced recording (P = 0.019). The earliest filtered onset in any lead preceeded the earliest surface activity by greater than 12 msec, in 6 VT versus one paced recording (P = 0.019). Early diastolic (post-QRS) potentials were also longer in VT than pacing (49 +/- 40 versus 5 +/- 20, P = 0.001) and exceeded 38 msec in seven of the VTs but none of the paced recordings (P = 0.001).(ABSTRACT TRUNCATED AT 250 WORDS)     

Automatic detection of ECG electrode misplacement: a tale of two algorithms

Artifacts in an electrocardiogram (ECG) due to electrode misplacement can lead to wrong diagnoses. Various computer methods have been developed for automatic detection of electrode misplacement. Here we reviewed and compared the performance of two algorithms with the highest accuracies on several databases from PhysioNet. These algorithms were implemented into four models. For clean ECG records with clearly distinguishable waves, the best model produced excellent accuracies (> = 98.4%) for all misplacements except the LA/LL interchange (87.4%). However, the accuracies were significantly lower for records with noise and arrhythmias. Moreover, when the algorithms were tested on a database that was independent from the training database, the accuracies may be poor. For the worst scenario, the best accuracies for different types of misplacements ranged from 36.1% to 78.4%. A large number of ECGs of various qualities and pathological conditions are collected every day. To improve the quality of health care, the results of this paper call for more robust and accurate algorithms for automatic detection of electrode misplacement, which should be developed and tested using a database of extensive ECG records.     

Evaluation of various electrocardiographic criteria for left ventricular hypertrophy in patients with stable angina pectoris: influence of using modified limb electrodes

Background: Left ventricular hypertrophy (LVH) in coronary heart disease is associated with poor prognosis. Electrocardiography (ECG) criteria for LVH, when using ECG with modified limb electrode positions, has not been validated in patients with angina pectoris. Methods: Echocardiography and resting ECGs with modified limb electrode positions, i.e. with the limb leads placed on the abdomen instead of the extremities, were registered from 468 patients (295 men) with stable angina pectoris. To evaluate the influence of using modified limb electrode positions, ECGs with standard and modified limb electrode positions were compared in a control group consisting of 50 other patients. Results: The ECG criteria for LVH according to the Perugia score, the Minnesota code and Romhilt & Estes reached the highest sensitivity values, 27–31% in men and 24–38% in women, while the sensitivities of different Cornell criteria were as low as 6–10% in men and 19–29% in women. In the control group, the R‐ and S‐wave amplitudes of the precordial leads were only slightly changed, as expected, whereas those of the limb leads changed considerably. Based on these results, we corrected aVL in the main study, which increased the sensitivity of the Cornell voltage criteria from 15 to 30%, while the specificity was maintained at 95%. Conclusions: ECGs registered with modified limb electrode positions can be used to detect LVH with traditional ECG criteria, but changes in the limb leads are considerable and influence the sensitivities.     

Diagnostic and prognostic importance of comparing the initial to the previous electrocardiogram in patients admitted for suspected acute myocardial infarction

Two hundred fifty-eight patients were admitted to the hospital for suspected acute myocardial infarction. Electrocardiograms recorded on admission (initial ECG) and the most recent available electrocardiogram recorded before admission (previous ECG) were compared to determine whether changes from the previous to initial ECG predicted acute myocardial infarction or complications of coronary artery disease. Initial ECGs were classed as either positive or negative, with positive indicating either infarction, injury, ischemia, strain, left ventricular hypertrophy, left bundle branch block, or paced rhythm. Negative ECGs were those that did not include any of the positive criteria. Positive and negative ECGs were subgrouped as showing change or no change from previous ECG. We found that patients with a negative initial ECG that had changed from the previous ECG had a 2.1 times greater risk for requiring interventions than those patients whose ECGs were unchanged. We also found that patients with a positive initial ECG that had changed from the previous ECG had a greater risk for interventions (2.0 times), complications (2.6 times), life-threatening complications (4.2 times), and acute myocardial infarction (6.6 times) than the sum of patients in all other ECG categories. We conclude that change is a useful predictor for interventions in patients with negative initial ECGs and a useful predictor for interventions, complications, and acute myocardial infarction in patients with positive initial ECGs.