Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms.
Andrew S TsengMichal Shelly-CohenItzhak Z AttiaPeter A NoseworthyPaul A FriedmanJae K OhFrancisco Lopez-JimenezPublished in: European heart journal. Digital health (2021)
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.
Keyphrases
- artificial intelligence
- machine learning
- deep learning
- big data
- aortic stenosis
- transcatheter aortic valve replacement
- ejection fraction
- aortic valve replacement
- transcatheter aortic valve implantation
- aortic valve
- left ventricular
- early onset
- palliative care
- coronary artery disease
- electronic health record
- heart rate variability
- atrial fibrillation