Overview

Echocardiography is central to risk stratification, diagnosis, and monitoring of cancer therapy-related cardiac dysfunction. AI-guided echocardiography has demonstrated high accuracy and reliability across diverse cardiac populations, yet had not previously been validated in a dedicated cohort of patients with cancer, a population presenting unique clinical challenges. This study addressed that gap directly.

 

Study Design

This study included retrospective and prospective cohorts from the cardio-oncology centre at Royal Brompton Hospital, totalling 384 patients (282 retrospective, 102 prospective). Mean age was 62 years, 61% were women, and breast cancer was the most common malignancy at 28.2%. All echocardiographic studies were analysed using Us2.ai, with standard sonographer-derived measurements as the reference. Primary outcome was level of agreement for 2D LVEF. Secondary outcomes included agreement for additional parameters, agreement between AI-derived 2D and 3D LVEF, temporal variability, interstudy repeatability, and time to study completion.

 

Key Results

  • Mean manual LVEF 58.15% vs. AI-derived LVEF 58.11%, with good agreement (ICC 0.774, 95% CI 0.729 to 0.812, bias 0.092, 95% LoA -10.9 to 11.0)
  • Comparison with 3D LVEF showed narrower limits of agreement (bias 0.04, ICC 0.8, 95% LoA -9.97 to 10.0)
  • Interstudy variability was significantly lower for AI repeated measurements than manual assessment (SEM 0.21 vs. 3.14, p<0.01)
  • Temporal variability was comparable between methods (SEM 3.97 vs. 5.93, p=0.112)
  • Study completion time reduced by 43.3% with the AI workflow (20.56 vs. 36.29 minutes)

 

Why This Matters

Patients undergoing cancer treatment require frequent serial cardiac monitoring to detect CTRCD early. Variability in manual measurements adds complexity to this process, and time pressures in busy oncology settings create real barriers to comprehensive assessment. This study demonstrates that Us2.ai delivers accurate and reproducible LVEF assessment in cancer patients while substantially reducing the time required to complete each study, making comprehensive cardiac monitoring more practical at scale in cardio-oncology settings.

 

Conclusion

In a large real-world cardio-oncology cohort, AI-guided echocardiography demonstrated strong agreement with conventional echocardiography for LVEF, improved interstudy repeatability, and a significant reduction in study completion time, supporting integration into routine cardio-oncology workflows.

 

 

Andres, M. S., Maharajan, V., Llamedo, C., Cervantes, J., Ohri, S., Nazir, S., Khattar, R., & Lyon, A. R. (2026, August 28-31). Validation of a deep learning-based workflow for the interpretation of the echocardiogram in a cardio-oncology population [ePoster presentation]. ESC Congress 2026, Munich, Germany.