AI Algorithm for Detecting Cardiac Amyloidosis
Author Information
Author(s): Duffy Grant, Oikonomou Evan, Hourmozdi Jonathan, Usuku Hiroki, Patel Jigesh, Stern Lily, Goto Shinichi, Tsujita Kenichi, Khera Rohan, Ahmad Faraz S., Ouyang David
Primary Institution: Cold Spring Harbor Laboratory
Hypothesis
Can a computer vision detection algorithm improve the diagnosis of cardiac amyloidosis across multiple international sites?
Conclusion
The AI algorithm EchoNet-LVH can help diagnose cardiac amyloidosis more accurately and earlier.
Supporting Evidence
- EchoNet-LVH achieved an AUC of 0.896, indicating good discrimination ability.
- The algorithm showed a specificity of 0.988, meaning it rarely misidentified healthy individuals as having the disease.
- Sensitivity was 0.644, suggesting it correctly identified some patients with cardiac amyloidosis.
Takeaway
This study shows that a computer program can help doctors find a rare heart disease more easily.
Methodology
A multi-site retrospective case-control study using a deep learning algorithm to analyze echocardiogram videos.
Limitations
The study may not cover all demographics or settings, and further research is needed to confirm the findings.
Statistical Information
Confidence Interval
95% CI 0.875 – 0.916
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website