Abstract
Background: Cardiotoxicity is a major limitation of chemotherapy and radiotherapy for thoracic and systemic cancers, contributing significantly to morbidity and mortality among survivors. Early prediction and prevention are critical to balance oncologic efficacy with cardiovascular safety. Artificial intelligence (AI) offers powerful tools to improve risk stratification, enable earlier detection of subclinical injury, and guide treatment planning in cardio-oncology. Methods: We performed a comprehensive review of the literature on AI applications for cancer therapy-related cardiotoxicity. Evidence was identified from PubMed, Scopus, and Web of Science, focusing on electrocardiography, biomarkers, proteomics, extracellular vesicles, genomics, advanced imaging (echocardiography, cardiac magnetic resonance, computed tomography, nuclear imaging), and radiotherapy dose modeling (dosiomics). Translational insights from animal models and in vitro systems were also included. Methodological quality was appraised with reference to TRIPOD-AI, PROBAST-AI, and CLAIM standards. Results: AI applications span multiple domains. Machine learning models integrating biomarkers, exosomes, and extracellular vesicles show promise for noninvasive early detection. Deep learning enables automated analysis of echocardiographic strain and cardiac MRI mapping, while radiomics and dosiomics approaches combine imaging with cardiac substructure dose maps to predict and prevent late radiation-induced injury. Preclinical studies demonstrate AI-driven advances in small-animal imaging, histopathology quantification, and multi-omics data integration, supporting the discovery of translational biomarkers. Despite encouraging performance, most models remain limited by small cohorts, methodological heterogeneity, and scarce external validation. Conclusions: AI has the potential to transform cardio-oncology by shifting from reactive detection to proactive prevention of cardiotoxicity. Future research should prioritize multimodal integration, harmonized multicenter datasets, prospective validation, and guideline-based clinical trials. As emerging data are incorporated, the field is expanding rapidly-dynamic, complex, and evolving.