Frédéric Kornreich

Body surface potential mapping of ST segment changes in acute myocardial infarction. Implications for ECG enrollment criteria for thrombolytic therapy.

BackgroundSeveral large, randomized clinical trials have shown that early thrombolytic therapy substantially reduces early mortality after acute myocardial infarction (MI). In most trials, eligibility criteria include typical chest pain and diagnostic ST segment elevation in two or more contiguous leads of the standard 12-lead ECG. Unfortunately, large areas of the thoracic surface are left unexplored by the standard electrode positions. As a consequence, acute MI patients with ST elevation in regions not interrogated by the conventional electrodes may not receive reperfusion therapy and its attendant benefits. Methods and ResultsThe present study compares 120-lead body surface potential map (BSPM) data from 131 patients with acute MI and 159 normal control subjects (N). The MI population was stratified according to the location of ventricular wall motion abnormalities evidenced by radionuclide imaging into 76 patients with anterior MI (AMI), 32 patients with inferior MI (IMI), and 23 patients with posterior MI (PMI). BSPM were recorded within 24 hours of admission. Group mean BSPM of the ST segment were obtained for N, AMI, IMI, and PMI by sampling the time-normalized ST-T waveform at 18 equal intervals and averaging the voltages at each electrode site over the first five of these 18 ST-T time instants. Corresponding discriminant maps were also computed for each pairwise comparison (AMI versus N, IMI versus N, and PMI versus N) by subtracting the normal group mean voltages from each MI group mean voltages and by further dividing each resulting difference by the composite standard deviation calculated from the pooled groups. Discriminant analysis for each bigroup classification was also performed using as measurements the ST magnitudes in 120 electrode sites from each individual. Finally, the number of patients in each MI group with ST changes outside the 95% normal range was calculated for each electrode position. The following results were obtained: 1) In each MI group, ST depression departs more significantly from normal values than ST elevation. 2) The most significant ST changes (both ST elevation and ST depression) are observed in IMI, the least significant in AMI. 3) For each pairwise comparison, measurements from two lead sites are entered into the stepwise discriminant procedure: the first measurement is ST depression, the second ST elevation. Classification rates are 82% for AMI, 93% for PMI, and 100o for IMI at a specificity level of 95%. 4) From the six leads selected for optimal classification of the three MI groups, five are outside the area sampled by the conventional precordial electrodes. 5) The use of site-dependent thresholds for ST measurements based on 95% normal range yields the best compromise between sensitivity and specificity. A fixed threshold of 1 mm for ST elevation or ST depression produces increased sensitivity in AMI at the cost of marked loss in specificity and reduces sensitivity in both IMI and PMI with no benefit in specificity. ConclusionAnalysis of BSPM identifies areas on the torso where the most significant ST changes most fequently occur in acute MI. Two leads from areas with the most abnormal ST changes achieve optimal classification in each MI class. Of these six leads, five are outsidethe standard precordial lead positions. ST depression is the most potent discriminator for each MI group and contains information independent from ST elevation. Quantitative analysis of ST magnitude at each electrode site allows determination of best thresholds for ECG criteria. Appropriate selection of ECG leads may help remove inconsistencies in current ECG selection criteria and improve comparability of treatment results.

Identification of first acute Q wave and non-Q wave myocardial infarction by multivariate analysis of body surface potential maps.

BackgroundPatients with acute non-Q wave myocardial infarction (NQMI) appear to have more jeopardized residual myocardium at high risk for subsequent angina, reinfarction, or malignant arrhythmias than patients with acute Q wave myocardial infarction (QMI). Unfortunately, conventional electrocardiographic (ECG) criteria have limited utility in recognizing NQMI. Methods and ResultsThe present study combines the increased information content of body surface potential maps (BSPM) over the 12-lead ECG with the power of multivariate statistical procedures to identify a practical subset of leads that would allow improved diagnosis of NQMI. Discriminant analysis was performed on 120-lead data recorded simultaneously in 159 normal subjects and 308 patients with various types of myocardial infarction (MI) by using instantaneous voltages on time-normalized P, PR, QRS, and ST-T waveforms as well as the duration of these waveforms as features. Leads and features for optimal separation of 159 normals from 183 patients with recent or old QMI (group A) were selected. A total of six features from six torso sites accounted for a specificity of 96% and a sensitivity of 94%. All lead positions were outside the conventional electrode sites and selected features were voltages at mid-P, early and mid-QRS, and before and after the peak of the T wave. The discriminant function was then tested on 57 patients with acute NQMI (group B) and 68 patients with acute QMI (group C): Rates of correct classification were 91% and 93%, respectively. Because of the possible deterioration of the results caused by ST-T abnormalities also present in other clinical entities, a second classification model including an independent group of 116 patients with left ventricular hypertrophy (LVH) but without MI was developed. Two additional measurements were required, namely, P wave duration and a mid-QRS voltage on a lead located 10 cm below V,. Testing the model on both acute MI groups produced correct classification rates of 88% for acute NQMI and 93% for acute QMI. Group mean BSPM were plotted for the three MI groups at successive instants throughout the PQRST waveform. Typical patterns for each MI group were identified during PQRST by removing the corresponding normal variability at each electrode site from sequential MI maps. These standardized maps or discriminant maps provided information on the capability of each measurement at each electrode site and at each instant to separate each class of MI from the normal group (N). Striking similarities were observed between the three MI groups, particularly at mid-QRS and throughout ST-T. The closest resemblance was between acute NQMI and old QMI. Discriminant analysis was also performed on the 12-lead ECG: The first classification model (N versus MI) produced correct classification rates of 85% for acute QMI and 70% 7e for NQMI. With the second model (MI versus N or LVH), correct rates were 81% and 65%, respectively. ConclusionsDiagnosis of acute NQMI and QMI (also in the presence of LVH) can be improved substantially by appropriate selection of ECG leads and features. Comparison of discriminant maps from groups A, B, and C does not support the concept of acute NQMI as a distinct ECG entity but rather as a group with infarcts of smaller size. However, pathophysiological and clinical differences between acute NQMI and acute QMI influence long-term risks and may define different therapeutic approaches.

Identification of best electrocardiographic leads for diagnosing anterior and inferior myocardial infarction by statistical analysis of body surface potential maps

In view of the increasing interest in quantifying and modifying the size of myocardial infarction (MI), it is important to look for clinically practical subsets of electrocardiographic leads that allow the earliest and most accurate diagnosis of the presence and electrocardiographic type of MI. A practical approach is described, taking advantage of the increased information content of body surface potential maps over standard electrocardiographic techniques for facilitating clinical use of body surface potential maps for such a purpose. Multivariate analysis was performed on 120-lead electrocardiographic data, simultaneously recorded in 236 normal subjects, 114 patients with anterior MI and 144 patients with inferior MI, using as features instantaneous voltages on time-normalized QRS and ST-T waveforms. Leads and features for optimal separation of normal subjects from, respectively, anterior MI and inferior MI patients were selected. Features measured on leads originating from the upper left precordial area, lower midthoracic region and the back correctly identified 97% of anterior MI patients, with a specificity of 95%; in patients with inferior MI, features obtained from leads located in the lower left back, left leg, right subclavicular area, upper dorsal region and lower right chest correctly classified 94% of the group, with specificity kept at 95%. Most features were measured in early and mid-QRS, although very potent discriminators were found in the late portion of the T wave.(ABSTRACT TRUNCATED AT 250 WORDS)

The missing waveform and diagnostic information in the standard 12 lead electrocardiogram

Summary Body surface potential data from 562 subjects were used to determine optimal locations of electrocardiographic (ECG) leads for retrieving waveform and diagnostic information which is lost if ECG recording is limited to the standard 12 leads. The essence of this approach rests on the empirical evidence obtained by testing each of the 128 simultaneously recorded surface ECGs for the standard leads content using a least squares best fit procedure. The poorly fitted curves are considered as additional waveform sources; the number and the sites of this additional signal information are then determined. The total QRS-ST surface waveform information is found to be represented by the standard 12-lead ECG whose eight independent components were supplemented with four additional surface leads (“supplemented” standard lead system). As an alternative to this set of 12 waveforms, a lead system consisting of nine directly recorded, non-redundant surface ECGs was also designed and found capable of accounting for the same total waveform information (“substitute” 9-lead system). The diagnostic accuracy of both lead sets is compared with the standard 12 leads and the Frank leads in differential diagnosis between normal individuals and patients with myocardial infarction. Results both in the training group (277 subjects) and in the testing group (200 subjects) reveal that the diagnostic content of the Prank leads is less than that of the standard 12-leads. The Frank leads combined with the standard leads improve moderately the diagnostic information yield. The fraction of correctly classified subjects amounted to 79%, 83%, 88%, 95% and 95% for the Frank leads, the standard leads, the combined Frank leads and standard leads, the “supplemented” standard leads and the “substitute” 9-leads, respectively. The results obtained with the standard leads and compared with those achieved by other lead systems, both better and worse as far as the waveform information is concerned, indicate that the retrieval of more complete signal information undoubtedly improves diagnostic performance.

The Missing Waveform Information in the Orthogonal Electrocardiogram (Frank Leads): I. Where and How Can This Missing Waveform Information be Retrieved?

The orthogonal electrocardiogram (Frank leads) is a reasonable compromise between the demands of practicability and diagnostic accuracy. The accuracy has been enhanced by the use of computerized multivariate statistical procedures. The work of the last two or three decades, however, has shown that these leads account for only part of the total available surface information. In this paper, the missing waveform information with respect to the Frank leads is quantitatively and systematically determined and a method of retrieval is worked out.The essence of this method rests on the empirical evidence obtained by testing each of the 126 surface electrocardiograms for the XYZ content. The results are analyzed by a least squares, best fit procedure. The poorly fitted curves are considered as additional waveform sources. The number and the sites of the additional information are determined on a test group of 207 patients. The “total’ surface waveform information is found to be represented by nine surface leads. This set of waveforms is then tested on a control group consisting of 205 patients. An average resynthesis coefficient of 96% is reached. No further resynthesis is looked for because of the estimated noise level.Although the waveforms recorded at these sites represent the minimum number of unique building blocks capable of synthesizing any waveform on the body surface, they cannot predict the surface potential distribution at all surface points. This information is “total’ as far as statistical procedures-performing discriminant analysis on time-functions—are concerned. This “total’ waveform information consists of a set of electrocardiograms recorded in each individual at nine well-defined anatomical locations.

Qualitative and quantitative analysis of characteristic body surface potential map features in anterior and inferior myocardial infarction

Body surface potential maps were recorded from 120 electrode sites in 236 normal subjects and 258 patients with initial evidence of either anterior myocardial infarction (MI) or inferior MI to identify characteristic map patterns in both groups. After time normalization, averaged map distributions were displayed at 18 equal time intervals during both QRS and ST-T waveforms from the normal, anterior MI and inferior MI groups. At each time instant, the 120-point averaged normal map was subtracted in turn from the corresponding anterior and inferior MI maps; the resulting differences at each electrode site were divided by the pooled standard deviation and the obtained values (discriminant indexes), plotted as contour lines with 1 standard deviation increments, producing discriminant maps for each bi-group comparison. The most consistent discriminant patterns in 114 patients with anterior MI were observed in early QRS in the upper left anterior chest where abnormal negative voltages reflected loss of electric potentials while reciprocal changes were noticed in the lower back; by mid-QRS, both distributions had moved jointly and vertically, the former in the lower torso on the midsternal line, the latter in the upper back. In 144 patients with inferior MI, abnormal positive distributions were observed in early QRS in the upper back, followed later by excessive negative voltages in the inferior right anterior chest; at mid-QRS, both distributions had migrated horizontally, the former proceeding toward the upper anterior torso, the latter to the lower left dorsal area.(ABSTRACT TRUNCATED AT 250 WORDS)

Estimating ECG distributions from small numbers of leads

The utility of body surface potential mapping to improve interpretation of electrocardiographic information lies in the presentation of thoracic surface distributions to characterize underlying electrophysiology less ambiguously than that afforded by conventional electrocardiography. Localized cardiac disease or abnormal electrophysiology presents itself electrocardiographically on the body surface in a manner in which pattern plays an important role for identifying or characterizing these abnormalities. Thus, in myocardial infarction, transient myocardial ischemia, Wolff-Parkinson-White syndrome, or ventricular ectopy, observation of electrocardiographic potential patterns, their extrema, and their magnitudes permits localization and quantization of the abnormal activity. Conventional electrocardiography assesses pattern information incompletely and does not use information of distribution extrema locations or magnitudes. Thus, increases or decreases in the magnitudes of electrocardiographic features (ST-segment potential displacement, amplitude, or morphology of Q, R, S, or T waves) associated with changes in cardiac sources (ischemia, infarction, conduction abnormalities, etc.) as measured from fixed leads have a high likelihood of being misinterpreted if the distribution itself is changing. In this study, the authors demonstrate the utility of estimating distributions from small numbers of optimally selected leads, including conventional leads, to reduce uncertainty in the interpretation of electrocardiographic information. This issue is highly relevant when thresholds are used to detect significance of potential levels (exercise testing, detection of myocardial infarction, and continuous monitoring to assess ST-segment changes). Significance of this work lies in improved detection and characterization of abnormal electrophysiology using conventional or enhanced leadsets and methods to estimate thoracic potential distributions.

The Missing Waveform Information in the Orthogonal Electrocardiogram (Frank Leads) III. Computer Diagnosis of Angina Pectoris from "Maximal" QRS Surface Waveform Information At Rest

Nine surface electrocardiograms recorded on the thoracic surface at fixed and identical locations in 412 individuals were found to account for the maximal useful waveform information available in each individual. In other words, nine waveforms were capable of resynthesizing any waveform recorded on their thoracic surface. These nine waveforms were then submitted to multivariate statistical procedures and their diagnostic performance compared to the Frank leads on which the same procedures were applied. Before the data were fed into the computer, all waveforms were time-normalized and divided into eight equal parts, yielding 72 variables and 24 variables for the nine lead system and the Frank leads, respectively, for each individual.In this paper we attempted to discriminate between normal subjects and patients with documented angina pectoris (typical history and positive coronary angiography); myocardial infarction was excluded in these patients. Only the resting QRS complex was considered. With the 9-lead system, keeping the specificity (true negatives) at 90%, the sensitivity (true positives) is 76%; with the Frank leads, the same specificity yielded a sensitivity of 49%. The repeatability of the results on new independent controls was also found very satisfactory.The discrimination between patients with angina pectoris on one hand and left ventricular hypertrophy and myocardial infarction on the other hand resulted in a performance level of 89% and 87%, respectively, for the 9-lead system. A good correlation was also found between the extent of the coronary lesions (number of coronary vessels involved) and the fraction of correctly diagnosed patients.The present study concluded that the retrieval of more complete surface information results in an evident improvement of the diagnostic performance of electrocardiography.

II. Diagnosis of left ventricular hypertrophy and myocardial infarction from 'total' surface waveform information

Eight surface leads were found to account for the “total’ waveform information in 282 patients (145 normal subjects, 59 patients with LVH, and 78 with myocardial infarction). In each patient the Frank leads were also reconstructed. After time normalization of the eight leads and the XYZ leads and division of the QRS complex into eight equal parts, the resulting variables (64 in the eight lead system and 24 in the Frank lead system) were submitted to multivariate statistical procedures. In a first step, the variables which proved best for the differentiation between normal records and those from patients with LVH or myocardial infarction were selected through stepwise discriminant analysis. A discriminant function was then computed and applied to both pathological groups. The results clearly point up the superiority of the eight lead system. With the level of specificity kept constant at 95%, 91% of the patients with LVH and 95% of the patients with myocardial infarction were correctly classified. With the Frank leads 83% and 85%, respectively, were recognized. The reproducibility of the results also proved to be better with the eight lead system.

Map representation and diagnostic performance of the standard 12-lead ECG

The diagnostic information contained in the standard 12-lead electrocardiogram was assessed by comparing the classification results produced by the standard leads for various clinical settings, such as normal versus myocardial infarction or versus left ventricular hypertrophy to those achieved by 120-lead data or body surface potential maps (BSPMs). Separately, optimal signal leads were extracted from the BSPM by ranking all leads in function of their capability of reconstructing the BSPM. Ranking was achieved by deriving eigenvalues from the covariance matrix calculated from all leads and corresponding measurements. Thus, while comparing the results from the standard leads (diagnostic leads) to those from the original raw map data, a comparison was also performed with respect to the best signal leads, namely the four best and the eight best. From the results observed for all bi- and multigroup classifications, it appeared that the diagnostic yield of the 12 standard leads matched those obtained with a number of signal leads lying between 4 and 8. This indicated that a large overlap still existed between the leads composing the 12-lead ECG (in fact, only 8 independent leads). Another interesting observation resulted from this investigation: although classifiers (discriminating variables) used for classification were identical, whether they originated from the raw standard leads (derived from the raw maps) or from standard leads reconstructed with four or eight signal leads, reconstructed measurements performed better than original measurements. This paradox can be explained by looking at the respective F values. Indeed, since increased F values result from higher ratios between the difference of group means and the composite variance from the pooled groups, higher differences and/or smaller variances produce larger ratios and hence, better group separations.