Open Access
Open access
volume 2 issue 4 pages 561-567

Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms

Publication typeJournal Article
Publication date2021-07-14
scimago Q1
wos Q1
SJR1.494
CiteScore7.2
Impact factor4.4
ISSN26343916
Energy Engineering and Power Technology
Fuel Technology
Abstract
Aims

Spectrum bias can arise when a diagnostic test is derived from study populations with different disease spectra than the target population, resulting in poor generalizability. We used a real-world artificial intelligence (AI)-derived algorithm to detect severe aortic stenosis (AS) to experimentally assess the effect of spectrum bias on test performance.

Methods and results

All adult patients at the Mayo Clinic between 1 January 1989 and 30 September 2019 with transthoracic echocardiograms within 180 days after electrocardiogram (ECG) were identified. Two models were developed from two distinct patient cohorts: a whole-spectrum cohort comparing severe AS to any non-severe AS and an extreme-spectrum cohort comparing severe AS to no AS at all. Model performance was assessed. Overall, 258 607 patients had valid ECG and echocardiograms pairs. The area under the receiver operator curve was 0.87 and 0.91 for the whole-spectrum and extreme-spectrum models, respectively. Sensitivity and specificity for the whole-spectrum model was 80% and 81%, respectively, while for the extreme-spectrum model it was 84% and 84%, respectively. When applying the AI-ECG derived from the extreme-spectrum cohort to patients in the whole-spectrum cohort, the sensitivity, specificity, and area under the curve dropped to 83%, 73%, and 0.86, respectively.

Conclusion

While the algorithm performed robustly in identifying severe AS, this study shows that limiting datasets to clearly positive or negative labels leads to overestimation of test performance when testing an AI algorithm in the setting of classifying severe AS using ECG data. While the effect of the bias may be modest in this example, clinicians should be aware of the existence of such a bias in AI-derived algorithms.

Found 
Found 

Top-30

Journals

1
International Journal of Arrhythmia
1 publication, 7.69%
Journal of the American Society of Echocardiography
1 publication, 7.69%
Revista española de cardiología (English ed.)
1 publication, 7.69%
Revista Espanola de Cardiologia
1 publication, 7.69%
Canadian Journal of Cardiology
1 publication, 7.69%
Cochrane Database of Systematic Reviews
1 publication, 7.69%
BIO Web of Conferences
1 publication, 7.69%
Cureus
1 publication, 7.69%
BMJ Digital Health & AI
1 publication, 7.69%
International Journal of Cardiovascular Sciences
1 publication, 7.69%
Shiraz E Medical Journal
1 publication, 7.69%
Pharmaceutics
1 publication, 7.69%
1

Publishers

1
2
3
4
Elsevier
4 publications, 30.77%
Springer Nature
2 publications, 15.38%
American Society of Echocardiography
1 publication, 7.69%
Wiley
1 publication, 7.69%
EDP Sciences
1 publication, 7.69%
BMJ
1 publication, 7.69%
Sociedade Brasileira de Cardiologia
1 publication, 7.69%
Brieflands
1 publication, 7.69%
MDPI
1 publication, 7.69%
1
2
3
4
  • We do not take into account publications without a DOI.
  • Statistics recalculated weekly.

Are you a researcher?

Create a profile to get free access to personal recommendations for colleagues and new articles.
Metrics
13
Share
Cite this
GOST |
Cite this
GOST Copy
Tseng A. et al. Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms // European Heart Journal - Digital Health. 2021. Vol. 2. No. 4. pp. 561-567.
GOST all authors (up to 50) Copy
Tseng A., Shelly Cohen M., Attia I. Z., Noseworthy P. A., Friedman P. A., Oh J. H., Lopez-Jimenez F. Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms // European Heart Journal - Digital Health. 2021. Vol. 2. No. 4. pp. 561-567.
RIS |
Cite this
RIS Copy
TY - JOUR
DO - 10.1093/ehjdh/ztab061
UR - https://doi.org/10.1093/ehjdh/ztab061
TI - Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms
T2 - European Heart Journal - Digital Health
AU - Tseng, Andrew
AU - Shelly Cohen, Michal
AU - Attia, Itzhak Z
AU - Noseworthy, Peter A.
AU - Friedman, Paul A.
AU - Oh, J H
AU - Lopez-Jimenez, Francisco
PY - 2021
DA - 2021/07/14
PB - Oxford University Press
SP - 561-567
IS - 4
VL - 2
PMID - 36713099
SN - 2634-3916
ER -
BibTex |
Cite this
BibTex (up to 50 authors) Copy
@article{2021_Tseng,
author = {Andrew Tseng and Michal Shelly Cohen and Itzhak Z Attia and Peter A. Noseworthy and Paul A. Friedman and J H Oh and Francisco Lopez-Jimenez},
title = {Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms},
journal = {European Heart Journal - Digital Health},
year = {2021},
volume = {2},
publisher = {Oxford University Press},
month = {jul},
url = {https://doi.org/10.1093/ehjdh/ztab061},
number = {4},
pages = {561--567},
doi = {10.1093/ehjdh/ztab061}
}
MLA
Cite this
MLA Copy
Tseng, Andrew, et al. “Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms.” European Heart Journal - Digital Health, vol. 2, no. 4, Jul. 2021, pp. 561-567. https://doi.org/10.1093/ehjdh/ztab061.