Cardiovascular Abnormalities, Electrocardiogram
Conditions
Keywords
artificial intelligence, electrocardiography, diagnostic methods, telemedicine, normal electrocardiogram
Brief summary
This study will evaluate the performance of specialist physicians in interpreting normal electrocardiograms (ECGs) with and without the assistance of an artificial intelligence (AI) neural network. The primary aim is to determine whether AI support affects the rate of false-positive interpretations of normal tracings. Secondary aims include evaluating the time required for interpretation, the sensitivity for detecting abnormalities, and the effect on false positives in ECGs with major abnormalities according to the Minnesota Code system. All ECGs in the sample will be reviewed by a panel of three specialists, to determine the reference classification.
Interventions
DIAGNOSTIC_TESTAI-Assisted ECG Interpretation (AI-ECG)
Neural network-based AI software that analyzes ECG tracings and provides a classification as normal suggestion to the interpreting specialist.
DIAGNOSTIC_TESTSpecialist ECG Interpretation Without AI
Manual interpretation of ECGs by specialists without AI support, following standard diagnostic procedures
Sponsors
Uppsala University
Federal University of Minas Gerais
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
DIAGNOSTIC
Masking
NONE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* ECGs performed routinely by the Rede de Telemedicina de Minas Gerais (RTMG)
Exclusion criteria
* ECGs from patients younger than 18 years
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Precision (Positive Predictive Value) for detection of normal ECG tracings | One week | Precision (Positive Predictive Value) of detecting normal ECG by the physician or physician+model compared against the reference standard defined by a panel of three specialists. Precision (Positive Predictive Value) is defined by the number of true positive normal cases divided by all positive predictions. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Sensitivity, Specificity, Negative Predictive Value, and F1 score for detection of normal ECG tracings | One week | Accuracy evaluated by Sensitivity, Specificity, Negative Predictive Value, and F1 score of normal ECGs correctly identified by the physician or physician+model, in relation to a reference standard defined by a panel of three specialists. |
| ECGs with major abnormalities incorrectly classified as normal | One week | Ratio of ECGs with major abnormalities according to the Minnesota Code system among those incorrectly classified as normal by the physician or physician+model, in relation to a reference standard defined by a panel of three specialists. |
| Time of analysis for normal cases (seconds per case) | One week | Time required by the physician, or physician+model, to interpret normal ECGs, measured in seconds per case; the reference standard of normal cases defined by a panel of three specialists. |
Contacts
Primary ContactAntonio Luiz P. Ribeiro, MD, PhD
Backup ContactGabriela Miana M. Paixão, MD, PhD
Outcome results
None listed