Artificial intelligence analysis of the electrocardiogram for early identification of cancer therapy related cardiotoxicity: a systematic review and meta-analysis

Abstract

Cancer therapy-related cardiac dysfunction (CTRCD) is a common, often subclinical complication of cancer treatment, and current imaging-based surveillance is resource-intensive. Artificial intelligence (AI)-enabled ECG analysis offers a scalable approach for early cardiotoxicity risk prediction, but supporting evidence remains unsynthesized. To evaluate the pooled sensitivity, specificity, and predictive value of AI-enabled ECG for early CTRCD identification through a systematic review and meta-analysis PubMed, Embase, and Cochrane Central were searched for AI-based ECG studies on early CTRCD detection. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were pooled using random-effects models, with 95% of confidence interval (CI). Performance summarized using a summary receiver operating characteristic (SROC) curve and area under the curve (AUC) via parametric bootstrapping. Three studies comprising 5,153 anthracycline-treated patients were included, 66.2% female with a mean age 60.7 years. Mean follow-up ranging from 1.5 to 9.5 years. AI-based ECG analysis showed a pooled sensitivity of 82.1% (95% CI, 58.0-93.9%) and specificity of 87.7% (95% CI, 82.5-91.4%). The pooled NPV was 99% (95% CI, 98-99.5%), whereas the pooled PPV was 22.4% (95% CI, 8.7-46.7%). Diagnostic performance was favorable, with an SROC AUC of 0.91 (95% CI, 0.75-0.96). AI-enabled ECG analysis demonstrates encouraging overall diagnostic accuracy for early identification of CTRCD, with high specificity and NPV. Larger studies are needed to validate these results and to clarify the role of AI-ECG as a complementary tool within cardio-oncology surveillance strategies.Graphical abstract  SROC curve