Article | SEP 15, 2025
Detecting cardiac amyloidosis (CA) early can be challenging, as its signs often overlap with other forms of heart failure. Presented at ASE 2025, this multicenter study validates Us2.ai’s fully automated pattern recognition echocardiography approach, and shows that it can accurately identify CA using just a single apical four-chamber view.
The study included 1,142 patients across 10 centers, all of whom underwent comprehensive guideline-directed testing to confirm CA. It also assessed performance across CA subtypes: transthyretin (ATTR-CA) and light chain (AL-CA), and compared results with established multiparametric echocardiographic scores.
Key Findings:
| Metric / Subgroup | Pattern Recognition Model |
| Cohort | 787 confirmed CA (69%), 229 ATTR-CA negative (20%), 126 non-cardiac AL (11%) |
| AUC | 0.87 |
| Sensitivity | 74% overall 82% for wall thickness >12mm |
| Specificity | 93% overall Similar for wall thickness >12mm |
| Accuracy | 80% overall 83% for wall thickness >12mm |
| Feasibility | 89% overall 86% in AL-CA |
| Performance Notes | Significantly outperformed IWT in ATTR-CA; similar accuracy to Systemic AL score in AL-CA but higher feasibility |
