The American Journal of Cardiology
Artificial intelligence better flags structural heart disease
The model’s score for distinguishing disease from no disease was 0.80, versus 0.75 for prediction using clinical information and standard heart-tracing findings.The model’s score for distinguishing disease from no disease was 0.80, versus 0.75 for prediction using clinical information and standard heart-tracing findings.
Cross-sectional · 3,779 peopleStructural heart disease
Who was studied3,779 adults undergoing public-system cardiovascular screening
What changedAI scored 0.80 versus 0.75 for conventional prediction when distinguishing structural heart disease from no disease
The catchThe AI correctly classified only 50.4% of people without disease as negative