Diagnosis of Cardiac Amyloidosis on Echocardiography Using Artificial Intelligence - PubMed
3 months ago
- #echocardiography
- #artificial intelligence
- #cardiac amyloidosis
- Artificial intelligence (AI) improves diagnosis of cardiac amyloidosis (CA) on echocardiography by addressing imaging overlaps with other hypertrophic phenotypes.
- The study involved 5776 patients (2756 with CA and 3020 controls) from diverse global cohorts including the UK, Taiwan, the US, and Japan.
- AI-derived multiparametric echocardiographic scoring achieved accuracies of 79.5% in the US cohort and 79.7% in the Japanese cohort.
- A deep-learning model demonstrated higher accuracies: 96.2% in internal validation and 95.8% in internal test sets.
- External validation showed the deep-learning model's accuracies of 87.5% in the US and 88.4% in Japan, outperforming the multiparametric score.
- The deep-learning model effectively discriminated CA from other hypertrophic conditions like hypertension, hypertrophic cardiomyopathy, aortic stenosis, and chronic kidney disease.
- The deep-learning model classified more patients accurately than the AI-derived multiparametric score, with superior diagnostic accuracy (AUC 0.93 vs. 0.88).
- Both AI approaches accurately identify CA in diverse populations, with the deep-learning model offering better performance.