Overview

Cardio-oncology is a resource-intensive discipline that demands accuracy and reproducibility in serial echocardiographic examinations. AI-enhanced echocardiographic analyses have shown reliable results in various populations, but the value of AI-echo in detecting cancer therapy-related cardiac dysfunction in a real-world breast cancer setting had not been established. This study from the CARE trial evaluated the feasibility and agreement of Us2.ai for CTRCD classification and longitudinal trajectory assessment.

 

Study Design

CARE is a prospective observational study of women with breast cancer scheduled for neoadjuvant anthracycline therapy and/or HER2-targeted therapy. Echocardiographic examinations were performed before, during, and after completion of scheduled treatment in 515 unselected patients by experienced cardiologists and sonographers. All examinations were also analysed using Us2.ai. CTRCD was defined as mild, moderate, and severe according to the 2022 ESC Cardio-oncology guidelines. Agreement between manual and AI-derived classifications was assessed, and longitudinal trajectories in LVEF and GLS were analysed using linear mixed-effects models.

 

Key Results

  • 2,225 paired echocardiographic examinations from 515 patients were included
  • AI-based measurements were feasible in 91% of examinations for LVEF and 93% for GLS
  • Overall CTRCD incidence was 24%: 19% mild, 5% moderate, 1% severe
  • Mean bias for GLS was -1.3% (r=0.53) and for LVEF was -2.1% (r=0.57)
  • Change-from-baseline bias was near zero for both LVEF (-0.10 pp) and GLS (0.11 pp)
  • Raw agreement between AI and manual CTRCD classification was 78.2%
  • Sensitivity 40% and specificity 89% for AI detection of any CTRCD assessed manually; NPV 91%
  • Longitudinal changes in LVEF and GLS showed parallel patterns with no significant time by method interactions

 

Why This Matters

The serial nature of cardiac monitoring in breast cancer patients, often across many visits and time points, makes consistency and reproducibility critical. This study demonstrates that Us2.ai reliably tracks longitudinal changes in cardiac function alongside manual assessments, with near-zero change-from-baseline bias. The high NPV of 91% supports the potential utility of AI to rule out CTRCD efficiently, reducing the burden of manual review while maintaining clinical safety in a high-volume monitoring setting.

 

Conclusion

In a large real-world breast cancer cohort, AI-enhanced echocardiography reliably reproduced longitudinal trajectories and ruled out CTRCD. These findings support the potential role of Us2.ai in improving precision and workflow efficiency in cardio-oncology practice.

 

 

Orstad, E., Myhre, P. L., Jakobsen, V. V., Gulati, G., Nerdrum, T., Schirmer, H., Geisler, J., Steine, K., & Omland, T. (2026, August 28-31). Agreement between manual and AI-enhanced echocardiographic assessment of cancer therapy-related cardiac dysfunction in patients with early breast cancer - data from the CARE study [ePoster presentation]. ESC Congress 2026, Munich, Germany.