Automated serial ECG comparison improves computerized interpretation of 12-lead ECG

695Poster Session 1 / Journal of Electrocardiology 45 (2012) 693–696

Corresponding 12-lead ECGs were simulated for the 20 combinations(Table). The ECGs were classified into strict LBBB (terminal negativedeflection in V1, QRS duration ≥140ms with mid-QRS notching),LVH (Sokolow–Lyon or Cornell criteria) or no confounders (non-LVH,non-strict LBBB). ECGs meeting conventional LBBB criteria (terminalnegative deflection in V1, QRS duration ≥120ms and notched R wavein leads I, aVL, V5, V6) were also noted.

Results: Complete LBBB was diagnosed correctly by strict and conven-tional LBBB criteria in all cases (normal anatomy, LV dilation and LVH).However, conventional ECG LBBB criteria falsely diagnosed LBBB in allcases with combined LVH+LV dilation and the case with 10-mm LVdilation combined with 18-ms LV activation delay.

Table12-Lead ECG diagnosis of simulated ECGs.

Simulated conduction type

Normal IncompleteLBBB(6-ms delay)

IncompleteLBBB(12-ms delay)

IncompleteLBBB(18-ms delay)

CompleteLBBB

Anatomicalmodel

Normal Noconfounders

Noconfounders

Noconfounders

Noconfounders

StrictLBBB(conventionalLBBB)

LV dilated5mm

Noconfounders

Noconfounders

ECG LVH ECG LVH StrictLBBB(conventionalLBBB)

LV dilated10mm

Noconfounders

Noconfounders

Noconfounders

Noconfounders(conventionalLBBB)

StrictLBBB(conventionalLBBB)

LVH+LVdilated10mm

ECG LVH(conventionalLBBB)

ECG LVH(conventionalLBBB)

ECG LVH(conventionalLBBB)

ECGLVH(conventionalLBBB)

StrictLBBB(conventionalLBBB)

Conclusions: This study supports the superiority of the strict (overconventional) LBBB criteria to correctly diagnose LBBB in the presenceof LV dilation and hypertrophy. Utilizing these strict criteria couldpotentially improve LBBB diagnostic specificity.

http://dx.doi.org/10.1016/j.jelectrocard.2012.08.031

C-reactive protein level is linked to T-wave axis deviation in a healthyItalian population: results from the MOLI-SANI projectMartino Vaglioa, Livia Ragob, Augusto Di Castelnuovob,Deodato Assanellic, Fabio Badilinia, Massimo Salvettic,Maria Benedetta Donatib, Giovanni de Gaetanob, Licia IacoviellobaAMPS LLC, New York, NY, USAbLaboratory of Genetic and Environmental Epidemiology. Fondazione diRicerca e Cura, “Giovanni Paolo II”, Catholic University, Campobasso, ItalycDepartment of Internal–Sport Medicine, University of Brescia, Brescia, Italy

Background:We aimed at investigating the link between C-reactive proteinlevel and ECG T-wave axis deviation in the Italian adult population.

Methods: A total of 10,655 women (mean age 55±12years) and 9472 men(mean age 56±12years) were analyzed from the cohort of the Moli-saniproject, a database of randomly recruited healthy adults (age N35years) fromthe general population of Molise, a central region of Italy that includescollection of standard 12-lead resting electrocardiogram.The goal of this study is to correlate the level of C-reactive protein with thedegree of T-wave axis deviation (TDev) defined as the rotation of the T-wave in the frontal plane and computed by a proprietary algorithm (CalECG/Bravo, AMPS LLC, NY). TDev was categorized in three separate groups:normal (≥15° to ≤75°), borderline (N75° to ≤105° or b15° to ≥−15°) andabnormal (b−15° to ≥−180° or N105° to ≤180°).

Results: Seventy-four percent, 24% and 2% of men had normal, borderlineand abnormal TDev, respectively, while women had 80%, 18% and 2%,respectively. On both genders an increase on C-reactive protein level waslinked with a worsening of TDev.

Men with abnormal TDev had 32% higher C-reactive protein level than maleswith normal TDev (pb0.0001, p value always adjusted for age, smoke, socialstatus, physical activity, CRP). Similarly for the female population withabnormal TDev, 25% higher C-reactive protein was found (pb0.0001).T-wave axis deviation was also strongly linked to BMI: men and womenwith abnormal TDev had 6% and 8% higher BMI than gender-matchedpopulation with normal TDev (pb0.0001). For both males and females, thepercentage of obese with abnormal TDev was twofold than for subjects withnormal BMI. The percentage of obese males with normal TDev was 61%compared with 89% for males with normal BMI. Similarly for women thetwo percentages were 73% and 90%, respectively.Conclusion: Our findings suggest that ECG monitoring could help tohighlight worsening of the subjects condition, prior an increase of C-reactiveprotein level. The effect of increased C-reactive protein level needs to beevaluated on other ECG and non-ECG markers.

http://dx.doi.org/10.1016/j.jelectrocard.2012.08.032

Automated serial ECG comparison improves computerizedinterpretation of 12-lead ECGRichard E. Gregga, Mario Arietb, Daniel C. Delucab,Cheng-hao Simon Chiena, Eric D. HelfenbeinaaAdvanced Algorithm Research Center, Philips Healthcare, Andover, MA, USAbDepartment of Medicine, University of Florida, Gainesville, FL, USA

Background: Interpretation of a patient's 12-lead ECG frequently involvescomparison to a previously recorded ECG. Automated serial ECGcomparison can be helpful not only to note significant ECG changes butalso to improve the single-ECG interpretation. Corrections from the previousECG are carried forward by the serial comparison algorithm whenmeasurements do not change significantly.Methods: A sample of patients from three hospitals were collected with twoor more 12-lead ECGs from each patient. There were 233 serial comparisonsfrom 143 patients. Forty-one percent of patients had two ECGs and 59% ofpatients had more than two ECGs. ECGs were taken from a difficultpopulation as measured by ECG abnormalities, 197/233 abnormal, 11/233borderline, 14/233 otherwise-normal and 11/233 normal. ECGs wereprocessed with the Philips DXL algorithm and then in time order for eachpatient with the Philips serial comparison algorithm. To measure accuracy ofinterpretation and serial change, an expert cardiologist corrected the ECGs instages. The first ECG was corrected and used as the reference for the secondECG. The second ECG was then corrected and used as the reference for thethird ECG and so on. At each stage, the serial comparison algorithmcompared an unedited ECG to an earlier edited ECG. Interpretation accuracywas measured by comparing the algorithm to the cardiologist on a statement-by-statement basis. The effect of serial comparison was measured by the sumof interpretive statement mismatches between the algorithm and thecardiologist. Statement mismatches were measured in two ways, (1) exactmatch and (2) match within the same diagnostic category.Results: The cardiologist used 910 statements over 233 ECGs for an averagenumber of 3.9 statements per ECG and a mode of 4 statements. Whenautomated serial comparison was used, the total number of exact statementmismatches decreased by 53% and the total same-category statementmismatches decreased by 54%.Conclusion: Automated serial comparison improves interpretation accuracyin addition to its main role of noting differences between ECGs.

http://dx.doi.org/10.1016/j.jelectrocard.2012.08.033

T-wave alternans: lessons learned from a biophysical ECG modelRoberto Sassia Luca T. MainardibaDipartimento di Tecnologie dell'Informazione, Università degli Studi diMilano,Crema, ItalybDipartimento di Bioingegneria, Politecnico di Milano, Milan, Italy

T-wave alternans (TWA) is an alteration of the ECG T-wave which repeats(alternates) every other beat. An alternating pattern has been also observed at