Computer interpretation of pediatric orthogonal electrocardiograms: statistical and deterministic classification methods.

Brohet, Christian; Robert, Annie; Derwael, C.; Fesler, Robert; Stijns, M.; Vliers, André; Braasseur, La.

doi:10.1161/01.cir.70.2.255

Cited by 19 publications

(7 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Quantitative patient tracking using the output of multidimensional models is not performed routinely for ECGs or even echos, in part because of long-standing fears that it might obscure the diagnostic process (52,53). With the current wide-spread availability of digital data, we strongly be-lieve such concerns should be revisited, both for the benefit of the physician and patient.…”

Section: Discussionmentioning

confidence: 99%

Automated and Interpretable Patient ECG Profiles for Disease Detection, Tracking, and Discovery

Tison

Zhang

Delling

et al. 2019

Circ: Cardiovascular Quality and Outcomes

122

106

View full text Add to dashboard Cite

Background:The electrocardiogram or ECG has been in use for over 100 years and remains the most widely performed diagnostic test for characterization of cardiac structure and electrical activity. Remarkably, current approaches to automated ECG interpretation originate from heuristics devised over 40 years ago. textbfObjective: We hypothesized that parallel advances in computing power, innovations in machine learning algorithms, and availability of large-scale digitized ECG data would enable extending the utility of the ECG beyond its current limitations, while at the same time preserving interpretability, an attribute which remains critical to medical decisionmaking. Methods: We identified 36,186 ECGs from the UCSF database that were 1) in normal sinus rhythm and 2) would enable training of specific models for estimation of cardiac structure or function or detection of disease.We derived a novel model for ECG segmentation using convolutional neural networks (CNN) and Hidden Markov Models (HMM) and evaluated its output by comparing electrical interval estimates to 141,864 measurements produced during the clinical workflow. We built a 725-element patient-level ECG profile using downsampled ECG segmentation data and trained machine learning models to estimate left ventricular mass, left atrial volume, mitral annulus e' and to detect and track four diseases: pulmonary arterial hypertension (PAH), hypertrophic cardiomyopathy (HCM), cardiac amyloid (CA), and mitral valve prolapse (MVP). Results: CNN-HMM derived ECG segmentation agreed with clinical estimates, with median absolute deviations (MAD) as a fraction of observed value of 0.6% for heart rate, 3% for PR interval, 4% for QT interval, and 6% for QRS duration. Patient-level ECG profiles enabled quantitative estimates of left ventricular mass (MAD vs. echocardiogram of 16%) and mitral annulus e' velocity (MAD of 19%) with good discrimination in binary classification models of left ventricular hypertrophy and diastolic dysfunction [Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.87 and 0.84, respectively]. Models for disease detection ranged from AUROC of 0.94 for PAH, 0.91 for HCM, 0.86 for CA, and 0.77 for MVP. Top-ranked variables for all models included known ECG characteristics along with novel predictors of these traits/diseases. Furthermore, temporal variation in model-derived disease scores coincided with visual evolution of ECG morphologies for these features. Conclusion: Modern AI methods can extend the 12-lead ECG to quantitative and diagnostic applications well beyond its current uses. Moreover, careful selection of machine learning algorithms achieves the goal of automation and accuracy without compromising the transparency that is so fundamental to clinical care and scientific discovery. electrocardiogram | machine learning | diagnosis | tracking Correspondence:

show abstract

Section: Discussionmentioning

confidence: 99%

Automated and Interpretable Patient ECG Profiles for Disease Detection, Tracking, and Discovery

Tison

Zhang

Delling

et al. 2019

Circ: Cardiovascular Quality and Outcomes

122

106

View full text Add to dashboard Cite

show abstract

“…In the literature, a number of advantages and disadvantages of the above classification methods have been put forward [1,7,13,17,18]. Several items will be reviewed here and discussed.…”

Section: Comparison Of Methodsmentioning

confidence: 99%

“…Numerous studies have been conducted to evaluate the performance of ECG computer programs. In only a few studies, a comparison was made between heuristic and statistical programs [1,15,18,22]. Generalization of the results of these studies is difficult since only two or three programs were involved.…”

Section: Performancementioning

confidence: 99%

Classification Methods for Computerized Interpretation of the Electrocardiogram

Kors¹,

Bemmel²

1990

Methods Inf Med

View full text Add to dashboard Cite

Two methods for diagnostic classification of the electrocardiogram are described: a heuristic one and a statistical one. In the heuristic approach, the cardiologist provides the knowledge to construct a classifier, usually a decision tree. In the statistical approach, probability densities of diagnostic features are estimated from a learning set of ECGs and multivariate techniques are used to attain diagnostic classification. The relative merits of both approaches with respect to criteria selection, comprehensibility, flexibility, combined diseases, and performance are described. Optimization of heuristic classifiers is discussed. It is concluded that heuristic classifiers are more comprehensible than statistical ones; encounter less difficulties in dealing with combined categories; are flexible in the sense that new categories may readily be added or that existing ones may be refined stepwise. Statistical classifiers, on the other hand, are more easily adapted to another operating environment and require less involvement of cardiologists. Further research is needed to establish differences in performance between both methods. In relation to performance testing the issue is raised whether the ECG should be classified using as much prior information as possible, or whether it should be classified on itself, explicitly discarding information other than age and sex, while only afterwards other information will be used to reach a final diagnosis. Consequences of taking one of both positions are discussed.

show abstract

“…After the initial evaluation using the validated database as a training set [26], the pediatric program was tested on a group of 500 unselected cases whose VCGs were prospectively collected as part of the daily routine task and, again, documented by means of ECG-independent data [27]. Table 2 shows the results in terms of completely correct classifications by deterministic and statistical methods.…”

Section: Analysis Of Pediatric Vcgsmentioning

confidence: 99%

Methodology of ECG Interpretation in the Louvain Program

Brohet

Derwael²,

Robert³

et al. 1990

Methods Inf Med

View full text Add to dashboard Cite

The Louvain program performs the analysis and interpretation of the vectorcardiogram (VCG) to increase the clinical utility of ECG analysis. Among its original features, there are (1) a high-resolution vector-loop display for visual analysis, (2) quantitative analysis of the spatial VCG using age-sex stratified limits, (3) separate software for adult and pediatric series and (4) complementary deterministic and statistical methods of diagnostic classification. Using objective, ECG-independent evidence as a reference standard, the Louvain program has shown satisfactory levels of diagnostic accuracy in most basic categories. However, its usefulness is especially marked in “borderline” or “complex” situations, where the 12-lead ECG cannot provide a clear answer. It corresponds to the concept of “computer-assisted ECG interpretation” as opposed to “computer ECG analysis”.

show abstract

Computer interpretation of pediatric orthogonal electrocardiograms: statistical and deterministic classification methods.

Cited by 19 publications

References 19 publications

Automated and Interpretable Patient ECG Profiles for Disease Detection, Tracking, and Discovery

Automated and Interpretable Patient ECG Profiles for Disease Detection, Tracking, and Discovery

Classification Methods for Computerized Interpretation of the Electrocardiogram

Methodology of ECG Interpretation in the Louvain Program

Contact Info

Product

Resources

About