Heart failure & cardiomyopathiesDiagnostic or prognostic study
Electrocardiogram-Based Deep Learning to Prioritize Testing for Transthyretin Amyloid Cardiomyopathy
- Diagnostic accuracy study
- temporal and external validation
- multinational
Development: 28 174 electrocardiography (ECGs) from 11 291 patients (293 with ATTR-CM)
Authors' conclusion
Locally deployable AI applied to ECG images discriminated ATTR-CM across multinational retrospective and screening cohorts.
Study context
- Population
- Patients evaluated for ATTR-CM (confirmed by cardiac amyloid radionuclide imaging or biopsy); development at Yale New Haven Health, external multinational and screening cohortsinternal validation
- Participants
- 44,123 patients
- Outcome
- AUROC, sensitivity, specificity, PPV and NPV for ATTR-CM confirmed by CARI or biopsy
Performance
As reported in the abstract
AUROC
0.84 (95% CI, 0.79-0.89)
Sensitivity at prespecified threshold
0.72
Specificity at prespecified threshold
0.86
Further findings
Specificity stress test: features mimicking ATTR-CM (LV hypertrophy or severe AS without amyloid)
- AUROC
- 0.81 [95% CI, 0.75-0.86]
Patients referred for cardiac amyloid radionuclide imaging (CARI)
- AUROC
- 0.78 [95% CI, 0.73-0.84]
External validation (5 cohorts)
- AUROC range
- 0.78 to 0.89
Screening cohorts
- AUROC, SCAN-MP (n = 645)
- 0.76 (95% CI, 0.68-0.83)
- AUROC, CACTUS cohort (n = 251)
- 0.79 (95% CI, 0.65-0.93)
- AUROC, CACTUS cohort (n = 121)
- 0.91 (95% CI, 0.82-0.97)
Sequential AI-ECG then artificial intelligence (AI)-enabled echocardiography (exploratory)
- PPV increased
- 0.24 to 0.66
- Sensitivity reduced
- 0.84 to 0.68
- exploratory analysis
Original publication
Summarised from the published abstract; every number is checked against it