Heart failure & cardiomyopathies
Diagnosis of Cardiac Amyloidosis on Echocardiography Using Artificial Intelligence
Authors' conclusion
Both the multiparametric echocardiographic score computed from AI-derived measurements and the fully automated deep-learning model can accurately identify patients with CA in globally diverse cohorts, with the deep-learning model providing superior performance.
Study context
- Population
- Patients with cardiac amyloidosis and controls, from the UK, Taiwan, the US and JapanCardiac amyloidosis (CA), 2756, controls, 3020
- Participants
- 5,776 patients
- Index test or model
- Deep-learning model versus AI-derived multiparametric echocardiographic score for detection of CA
- Target condition
- cardiac amyloidosis (CA)
Methodological detailsexternal test data sets
- External test data sets
Performance
As reported in the abstract
| Test or model | Sensitivity | Specificity | Accuracy |
|---|---|---|---|
| AI-derived multiparametric echocardiographic score, United States cohort | 75.4% | 81.5% | 79.5% |
| AI-derived multiparametric echocardiographic score, Japan cohort | 81.6% | 78.1% | 79.7% |
| Deep-learning model, internal validation set | 96.8% | 95.7% | 96.2% |
| Deep-learning model, internal test set | 97.3% | 94.3% | 95.8% |
| Deep-learning model, external validation, United States | 86.6% | 87.9% | 87.5% |
| Deep-learning model, external validation, Japanese cohort | 92.3% | 85.3% | 88.4% |
Sensitivity
AI-derived multiparametric echocardiographic score, United States cohort75.4%
AI-derived multiparametric echocardiographic score, Japan cohort81.6%
Deep-learning model, internal validation set96.8%
Deep-learning model, internal test set97.3%
Deep-learning model, external validation, United States86.6%
Deep-learning model, external validation, Japanese cohort92.3%
Specificity
AI-derived multiparametric echocardiographic score, United States cohort81.5%
AI-derived multiparametric echocardiographic score, Japan cohort78.1%
Deep-learning model, internal validation set95.7%
Deep-learning model, internal test set94.3%
Deep-learning model, external validation, United States87.9%
Deep-learning model, external validation, Japanese cohort85.3%
Accuracy
AI-derived multiparametric echocardiographic score, United States cohort79.5%
AI-derived multiparametric echocardiographic score, Japan cohort79.7%
Deep-learning model, internal validation set96.2%
Deep-learning model, internal test set95.8%
Deep-learning model, external validation, United States87.5%
Deep-learning model, external validation, Japanese cohort88.4%
Further findings
AUC
| Outcome | AI-derived multiparametric echocardiographic score | Deep-learning model | P value |
|---|---|---|---|
| Deep-learning model vs AI-derived multiparametric echocardiographic score, diagnostic accuracy | 0.88 (95% CI 0.85–0.90) | 0.93 (95% CI 0.91–0.95) | P < 0.001 |
- Deep-learning model vs AI-derived multiparametric echocardiographic score, diagnostic accuracy: superior diagnostic accuracy with the deep-learning model; it classified a greater proportion of patients
Deep-learning model discrimination of CA from other hypertrophic phenocopies (subgroup analysis)
- CA versus hypertension, AUC
- 0.92 [95% CI, 0.91-0.94]
- CA versus hypertrophic cardiomyopathy, AUC
- 0.91 [95% CI, 0.87-0.94]
- CA versus aortic stenosis, AUC
- 0.93 [95% CI, 0.90-0.95]
- CA versus chronic kidney disease, AUC
- 0.93 [95% CI, 0.91-0.95]
Original publication
Diagnosis of Cardiac Amyloidosis on Echocardiography Using Artificial Intelligence. Circ Cardiovasc Imaging. Published 12 February 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →