Fully automated artificial intelligence–based echocardiographic analysis for global longitudinal strain monitoring and cancer therapy–related cardiac dysfunction detection in breast cancer patients

Abstract Aims Global longitudinal strain (GLS) is essential for the early detection of cancer therapy–related cardiac dysfunction (CTRCD). A fully automated echocardiographic analysis system using artificial intelligence (AI) may improve workflow efficiency in cardio-oncology. We sought to evaluate the feasibility and diagnostic performance of a fully automated AI-based echocardiographic system in breast cancer patients receiving cardiotoxic chemotherapy. Methods and Results In this prospective observational study, patients with breast cancer undergoing anthracyclines and/or HER2-targeted therapy between January 2022 and June 2025 were enrolled. Transthoracic echocardiography was performed at baseline and every 12 weeks. GLS was measured manually by two experts and automatically by a fully automated AI-based analysis system. A total of 92 patients (456 echocardiographic studies) were analysed. AI-derived GLS values were significantly lower than expert measurements (17.7 ± 2.9% vs. 18.4 ± 2.8%, P = 0.007). Correlation and agreement between the two methods were moderate (R = 0.64, intraclass correlation coefficient = 0.63). On linear mixed-effects modelling, longitudinal changes in GLS were not significantly different between methods (P = 0.72). GLS-based CTRCD was detected in 31.5% of patients by experts and 34.8% by AI (P = 0.58), with similar detection timing (P = 0.47). Diagnostic agreement was substantial (κ = 0.68, P < 0.001). Conclusion The fully automated AI-based echocardiographic system demonstrated acceptable agreement and diagnostic performance for GLS assessment and showed a similar ability to track temporal relative GLS changes and identify CTRCD. However, systemic underestimation of absolute GLS values may contribute to threshold-based classification discordance in borderline cases.