Bridging gaps in AI-driven cardio-oncology: global advances and Middle East perspectives Authors: Arif Albulushi, Hatem Al-Farhan Published: 2025 ABSTRACT BackgroundCardiotoxicity remains a critical barrier to effective cancer treatment, contributing significantly to morbidity and mortality in cancer survivors. Recent advancements (2020–2025) highlight artificial intelligence (AI) as a transformative tool in cardio-oncology, enhancing cardiotoxicity detection, personalized risk assessment, and patient management. Despite rapid technological progress, significant gaps remain in AI integration across different healthcare settings globally.MethodsThis narrative review systematically synthesizes literature published between 2020 and 2025, evaluating key applications of AI in cardio-oncology, including imaging modalities, predictive modeling, wearable technology, and clinical decision-support systems. Comparative analyses between high-income countries (HICs) and low-to-middle-income countries (LMICs), with an emphasis on the Middle East, were performed to illustrate global disparities and unique regional challenges.ResultsAI-enabled imaging technologies, particularly echocardiography and cardiac MRI, significantly improved the early detection and management of cardiotoxicity. Predictive algorithms integrating multimodal data demonstrated superior risk stratification accuracy over traditional methods. Wearable technologies combined with AI enabled real-time cardiac monitoring, demonstrating feasibility in diverse resource settings. Nonetheless, adoption barriers such as dataset biases, inadequate regulatory frameworks, prohibitive costs, ethical dilemmas, and limited digital infrastructure persist, disproportionately affecting LMICs.ConclusionTo harness the full potential of AI in cardio-oncology, strategic investments in collaborative international research, standardized regulatory frameworks, educational initiatives, and infrastructural support are urgently needed. Future research should prioritize equitable, inclusive AI solutions, validated through prospective trials and adapted specifically to underserved populations.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40959-025-00411-9. Introduction Advancements in cancer therapeutics over the past two decades have significantly improved survival rates, yet this progress has been accompanied by a substantial rise in cardiovascular (CV) complications, collectively termed cardiotoxicities. Cardiotoxicity in cancer patients can manifest as asymptomatic myocardial dysfunction, overt heart failure, thromboembolic events, arrhythmias, and vascular complications, posing a major challenge for both oncologists and cardiologists alike [1, 2]. Anthracyclines, HER2-targeted therapies, and immune checkpoint inhibitors remain among the most well-documented agents contributing to adverse CV outcomes such as cardiomyopathy and myocarditis [2, 3]. The field of cardio-oncology has emerged to address this complex overlap, aiming to improve cancer outcomes while minimizing CV harm through early detection, personalized risk prediction, and timely intervention [4, 5]. However, conventional diagnostic tools, such as routine echocardiography and serum biomarkers, often lack sufficient sensitivity to detect early subclinical myocardial injury, limiting their utility for proactive cardioprotection [6]. Recent breakthroughs from 2020 to 2025 highlight the transformative role of artificial intelligence (AI) as an adjunct in cardio-oncology practice. AI-powered algorithms now enhance image-based modalities like echocardiography and cardiac MRI, enable continuous monitoring through wearable devices, and integrate multimodal datasets to build robust predictive models for cardiotoxicity risk [7–9]. For instance, deep learning algorithms can detect subtle changes in left ventricular strain on echocardiography years before overt dysfunction manifests, outperforming human interpretation alone [7–10]. Despite its promise, global adoption of AI in cardio-oncology remains uneven. High-income countries (HICs) — notably the United States, Canada, and parts of Europe — have invested heavily in AI research, digital infrastructure, and regulatory sandboxes, facilitating faster implementation [11–13]. By contrast, low- and middle-income countries (LMICs), including many Middle Eastern and North African nations, still grapple with limited computational resources, lack of local validation cohorts, ethical concerns, and gaps in trained workforce capacity [14, 15]. Encouragingly, the Middle East region has made notable strides in digital health policy through national visions such as Saudi Arabia’s Vision 2030, Oman’s Health Vision 2040, and the UAE’s digital transformation agenda, each prioritizing AI integration in healthcare [15–17]. However, standardization, interoperability, and sustainable workforce development remain critical hurdles [17, 18]. Evidence suggests that cancer patients in the Middle East may experience a higher incidence of treatment-induced cardiotoxicity compared to Western cohorts, further underscoring the need for region-specific AI-enabled cardio-oncology programs [19, 20]. Recognizing these gaps, recent global consensus documents from the World Health Organization (WHO) and professional societies have emphasized equitable and ethical AI deployment, with a focus on improving data diversity, governance, and local capacity building [21–23]. The aim of this updated narrative review is to synthesize the latest (2020–2025) advances in AI applications within cardio-oncology, highlight persisting implementation barriers globally and regionally, and propose actionable strategies to bridge disparities — with a particular lens on Middle Eastern health systems. By consolidating current evidence, we hope to inform stakeholders on how to realistically harness AI to deliver safer, more equitable cardio-oncological care worldwide. AI-enhanced imaging and diagnostics Early detection of cardiotoxicity remains the cornerstone of effective cardio-oncology management, yet traditional imaging and biomarker strategies often fall short in identifying subclinical myocardial dysfunction before irreversible damage occurs [1, 4]. Artificial intelligence (AI) has emerged as a transformative tool to bridge this diagnostic gap by augmenting the sensitivity, reproducibility, and precision of cardiac imaging modalities. AI-enhanced echocardiography, particularly deep learning-based left ventricular (LV) strain analysis, has demonstrated superior performance in detecting subtle myocardial deformation changes compared to conventional strain quantification [7, 24]. Recent prospective studies show that AI-automated strain can flag early LV dysfunction in anthracycline-treated patients up to six months before standard ejection fraction declines, enabling timely cardioprotective interventions [25]. Beyond echocardiography and CMR, coronary computed tomography angiography (CCTA) and coronary artery calcium (CAC) scoring on non-gated chest CTs, often performed for oncologic staging, represent valuable opportunities for baseline cardiovascular risk stratification. AI-enhanced CAC quantification has been shown to predict future atherosclerotic burden and cardiotoxicity risk, enabling oncologists to proactively tailor surveillance and preventive therapies [26]. In parallel, artificial intelligence applied to electrocardiography (AI-ECG) has emerged as a low-cost, scalable tool in cardio-oncology. AI-ECG algorithms can detect subclinical left ventricular dysfunction, arrhythmogenic risk, and even predict the development of cardiomyopathy prior to overt ejection fraction decline, which may be particularly useful in resource-constrained regions. Taken together, multimodality AI-driven diagnostics—spanning echocardiography, CMR [8, 27], CT/CAC, and AI-ECG—not only enhance early detection but also provide prognostic insights into long-term cardiovascular outcomes in cancer survivors (Fig. 1). Fig. 1Workflow for AI-driven cardiotoxicity detection. This figure illustrates how AI algorithms integrate multimodal inputs (imaging, biomarkers, EMR data) to enhance cardiotoxicity detection and clinical decision-making. It was constructed based on concepts synthesized from studies cited in section "AI-enhanced imaging and diagnostics" and "AI-enabled clinical decision support and workflow integration" of the manuscript Workflow for AI-driven cardiotoxicity detection. This figure illustrates how AI algorithms integrate multimodal inputs (imaging, biomarkers, EMR data) to enhance cardiotoxicity detection and clinical decision-making. It was constructed based on concepts synthesized from studies cited in section "AI-enhanced imaging and diagnostics" and "AI-enabled clinical decision support and workflow integration" of the manuscript Collectively, AI-powered imaging innovations advance the precision of early cardiotoxicity detection and lay the groundwork for adaptive, patient-specific monitoring strategies that can optimize long-term cardiovascular and oncologic outcomes [28]. AI in remote monitoring and digital biomarkers Continuous monitoring is pivotal in mitigating cardiotoxicity among cancer patients, especially for detecting dynamic changes that static, clinic-based assessments may miss [5, 11]. Integrating artificial intelligence (AI) with wearable technologies has revolutionized how clinicians and patients track cardiovascular health in real time. AI-enabled smartwatches, biosensors, and patches now facilitate continuous surveillance of heart rate variability, arrhythmias, and blood pressure trends, providing early warning signals of impending decompensation [11, 29]. The COVID-19 pandemic further accelerated the uptake of remote monitoring solutions, demonstrating their feasibility for maintaining care continuity even in resource-constrained or geographically dispersed populations [5]. Recent studies confirm that AI-enhanced remote monitoring reduces unscheduled hospital visits and allows for proactive therapy adjustments in high-risk cardio-oncology cohorts [30]. A novel frontier in this domain is the use of passive smartphone-based digital biomarkers, such as voice signal changes and micro-motion sensing, to detect early signs of cardiac dysfunction. Bachtiger et al. showed that subtle alterations in voice acoustics, analyzed using machine learning, can predict decompensated heart failure with clinically acceptable accuracy [31]. These unobtrusive digital biomarkers complement wearable data and hold promise for expanding remote cardiotoxicity surveillance, especially in low-resource settings where access to advanced imaging remains limited [32]. Key challenges limiting AI adoption in this space are summarized in Table 1. Table 1Key challenges and ethical considerations in AI integration for cardio-oncology. This table categorizes challenges into four domains (data and bias, cost and access, ethical/privacy, and regulatory gaps), illustrates each with examples and supporting data from the literature, and provides proposed solutions based on global recommendationsCategoryKey ChallengesExamples/Supporting DataProposed SolutionsData & Bias- Underrepresentation in LMIC datasets- Algorithm bias, limited validation- Only 15% of models use African/South Asian data- Cardiotoxicity AI 20% less accurate in Black patients- Promote global data-sharing- Develop bias audit frameworksCost & Access- High development costs- Poor digital infrastructure− 85% LMIC hospitals lack AI readiness- Oncology AI $200–500 K per site- Subsidies for LMICs- Partner with telecoms for rural connectivityEthical & Privacy- Consent/privacy gaps- Low transparency- GDPR compliance challenges- Global consent templates- Open-source algorithm explainersRegulatory Gaps- No harmonized global AI regulation- WHO: few LMICs have AI policy- WHO/IMDRF-driven policy harmonization Table 1Key challenges and ethical considerations in AI integration for cardio-oncology. This table categorizes challenges into four domains (data and bias, cost and access, ethical/privacy, and regulatory gaps), illustrates each with examples and supporting data from the literature, and provides proposed solutions based on global recommendationsCategoryKey ChallengesExamples/Supporting DataProposed SolutionsData & Bias- Underrepresentation in LMIC datasets- Algorithm bias, limited validation- Only 15% of models use African/South Asian data- Cardiotoxicity AI 20% less accurate in Black patients- Promote global data-sharing- Develop bias audit frameworksCost & Access- High development costs- Poor digital infrastructure− 85% LMIC hospitals lack AI readiness- Oncology AI $200–500 K per site- Subsidies for LMICs- Partner with telecoms for rural connectivityEthical & Privacy- Consent/privacy gaps- Low transparency- GDPR compliance challenges- Global consent templates- Open-source algorithm explainersRegulatory Gaps- No harmonized global AI regulation- WHO: few LMICs have AI policy- WHO/IMDRF-driven policy harmonization Key challenges and ethical considerations in AI integration for cardio-oncology. This table categorizes challenges into four domains (data and bias, cost and access, ethical/privacy, and regulatory gaps), illustrates each with examples and supporting data from the literature, and provides proposed solutions based on global recommendations - Underrepresentation in LMIC datasets - Algorithm bias, limited validation - Only 15% of models use African/South Asian data - Cardiotoxicity AI 20% less accurate in Black patients - Promote global data-sharing - Develop bias audit frameworks - High development costs - Poor digital infrastructure − 85% LMIC hospitals lack AI readiness - Oncology AI $200–500 K per site - Subsidies for LMICs - Partner with telecoms for rural connectivity - Consent/privacy gaps - Low transparency - Global consent templates - Open-source algorithm explainers Together, AI-driven remote monitoring and emerging digital biomarkers represent powerful tools to bridge gaps in continuous cardio-oncology care, enhance patient engagement, and support timely clinical decision-making. AI-based predictive modeling and risk stratification Traditional risk stratification in cardio-oncology often relies on generalized clinical risk factors and cumulative chemotherapy dosing, which can miss individual variations in treatment response and susceptibility to cardiotoxicity [6, 8]. Artificial intelligence (AI)-driven predictive modeling now enables the integration of diverse patient data — including longitudinal clinical records, advanced imaging biomarkers, and genomic profiles — to deliver more granular, personalized risk assessments [8, 33]. Machine learning algorithms have demonstrated high accuracy in forecasting which patients are most likely to develop cardiotoxicity from specific agents such as trastuzumab and anthracyclines. Zhang et al. highlighted that predictive AI models incorporating imaging-derived strain parameters and dose intensity achieved over 90% accuracy for early identification of high-risk breast cancer patients [8]. A novel frontier is the integration of polygenic risk scores (PRS) with AI models. Combining common genetic variants with dynamic clinical and imaging data can uncover hidden susceptibilities that would otherwise go undetected [27, 34]. Abraham et al. showed that adding PRS to AI-powered predictive models improved risk stratification for anthracycline-induced cardiomyopathy by up to 20%, underscoring the synergy between genomics and machine learning in precision cardio-oncology [27]. These advancements lay the groundwork for precision prevention strategies — enabling clinicians to tailor surveillance intensity, modify treatment regimens, and implement cardioprotective interventions proactively based on an individual’s comprehensive risk landscape [35]. AI-enabled clinical decision support and workflow integration Integrating artificial intelligence (AI) into routine clinical workflows is vital to fully leverage its potential in cardio-oncology practice. AI-powered clinical decision support systems (CDSS) assist clinicians by providing real-time alerts for cardiotoxicity risk, recommending dose modifications, and guiding surveillance intervals, thereby reducing adverse events and optimizing treatment continuity [9, 33]. Recent pilot implementations of smart electronic medical record (EMR) plug-ins have shown that embedding AI algorithms directly into oncology and cardiology workflows significantly improves adherence to evidence-based guidelines and reduces manual workload [34, 36]. An emerging frontier is the adoption of large language models (LLMs) — such as GPT-like AI — as on-demand virtual clinical assistants. Chen et al. demonstrated that LLMs can process unstructured patient records, synthesize relevant history, and even suggest context-specific monitoring plans for complex cardio-oncology cases, all while supporting clinician oversight [36]. An additional emerging application is the integration of AI-guided point-of-care ultrasound (POCUS) into cardio-oncology workflows. Recent studies demonstrate that AI-assisted POCUS enables oncology nurses and non-expert providers to acquire and interpret focused echocardiographic images with high diagnostic accuracy, particularly for screening left ventricular dysfunction in patients receiving cardiotoxic therapies [37]. This approach not only improves workflow efficiency but also facilitates task-shifting in low-resource or high-volume oncology settings where cardiology specialists may be limited, as illustrated in Fig. 2. Fig. 2Ethical challenges in AI integration for cardio-oncology. This schematic summarizes key ethical domains—data privacy, transparency, bias, and accountability—derived from WHO and IMDRF guidelines and peer-reviewed literature. It highlights where ethical oversight must be prioritized when adopting AI tools in oncology and cardiology practice Ethical challenges in AI integration for cardio-oncology. This schematic summarizes key ethical domains—data privacy, transparency, bias, and accountability—derived from WHO and IMDRF guidelines and peer-reviewed literature. It highlights where ethical oversight must be prioritized when adopting AI tools in oncology and cardiology practice Incorporating nurse-led AI-POCUS programs into clinical decision support systems can enhance early detection of cardiotoxicity, streamline follow-up protocols, and expand access to high-quality cardiac imaging in underserved regions. Despite promising results, scaling CDSS and LLM tools requires robust data privacy safeguards, real-world validation, and clinician training to ensure safe, trustworthy implementation in diverse practice settings [38]. Regional implementation perspectives Despite rapid advancements in artificial intelligence (AI), significant disparities persist in its implementation across different regions. High-income countries (HICs) such as the United States, Canada, and Germany benefit from robust digital infrastructure, substantial funding, and specialized workforce capacity, enabling swift integration of AI-driven tools into cardio-oncology practice [7, 33]. These systems routinely leverage AI-powered imaging, remote monitoring, and decision support to enhance patient outcomes and streamline clinical workflows [39]. Conversely, low- and middle-income countries (LMICs) face multifaceted barriers including limited computational infrastructure, funding constraints, and gaps in digital health literacy [14, 40]. The World Health Organization (WHO) reports that over 85% of hospitals in LMICs lack sufficient digital health systems to support advanced AI integration, widening the global digital divide [14]. In the Middle East, regional initiatives such as Saudi Arabia’s Vision 2030, Qatar’s National AI Strategy, and the UAE’s AI Ministry reflect ambitious governmental commitment to digital health transformation [15, 18]. However, operational challenges remain. For example, Oman and other GCC countries continue to face practical hurdles including data interoperability, fragmented electronic health records, and a shortage of locally trained AI specialists [28, 41]. Recent work by Al Saidi et al. underscores the importance of region-specific frameworks and workforce development to accelerate AI adoption in Gulf cardio-oncology programs [28]. A comparative overview of Middle Eastern versus Western advancements is provided in Table 2. Table 2Regional advancements in AI and digital health: middle East vs. Western world. This comparative table was created using information synthesized from published studies, regional AI strategy documents (e.g., vision 2030, Oman vision 2040, Qatar National AI Strategy), and peer-reviewed cardio-oncology literature. It highlights current cancer burden, patterns of AI use, remote monitoring adoption, infrastructure, and innovation policy, and identifies future needs for each regionParameterMiddle EastWestern WorldFuture NeedsCancer Burden- Breast/colorectal upsurge- Cardiotoxicity ~ 25%- Prostate, lung prevalent- Cardiotoxicity 10–20%- Registry harmonizationAI Use- HER2 therapy monitoring (Oman/UAE)- Echo AI pilots- Predictive AI HER2/VEGF models- Broaden beyond pilotsRemote Monitoring- Slow cloud adoption- Internet gaps, costs- Wearables, telehealth routine- Affordable telehealthInfrastructure- Fragmented funding- NIH, Horizon Europe robust funding- Unified funding streamsInnovation- Vision 2030, Oman Vision 2050 push AI- UK AI Strategy supports health AI- Regional policy alignment Table 2Regional advancements in AI and digital health: middle East vs. Western world. This comparative table was created using information synthesized from published studies, regional AI strategy documents (e.g., vision 2030, Oman vision 2040, Qatar National AI Strategy), and peer-reviewed cardio-oncology literature. It highlights current cancer burden, patterns of AI use, remote monitoring adoption, infrastructure, and innovation policy, and identifies future needs for each regionParameterMiddle EastWestern WorldFuture NeedsCancer Burden- Breast/colorectal upsurge- Cardiotoxicity ~ 25%- Prostate, lung prevalent- Cardiotoxicity 10–20%- Registry harmonizationAI Use- HER2 therapy monitoring (Oman/UAE)- Echo AI pilots- Predictive AI HER2/VEGF models- Broaden beyond pilotsRemote Monitoring- Slow cloud adoption- Internet gaps, costs- Wearables, telehealth routine- Affordable telehealthInfrastructure- Fragmented funding- NIH, Horizon Europe robust funding- Unified funding streamsInnovation- Vision 2030, Oman Vision 2050 push AI- UK AI Strategy supports health AI- Regional policy alignment Regional advancements in AI and digital health: middle East vs. Western world. This comparative table was created using information synthesized from published studies, regional AI strategy documents (e.g., vision 2030, Oman vision 2040, Qatar National AI Strategy), and peer-reviewed cardio-oncology literature. It highlights current cancer burden, patterns of AI use, remote monitoring adoption, infrastructure, and innovation policy, and identifies future needs for each region - Breast/colorectal upsurge - Cardiotoxicity ~ 25% - Prostate, lung prevalent - Cardiotoxicity 10–20% - HER2 therapy monitoring (Oman/UAE) - Echo AI pilots - Slow cloud adoption - Internet gaps, costs Targeted investments, regional research collaborations, and harmonized governance policies will be critical to bridging these gaps and ensuring that AI benefits are equitably distributed across diverse populations. Frameworks, ethics, and governance The responsible integration of artificial intelligence (AI) in cardio-oncology requires robust ethical frameworks and clear governance structures to protect patient safety, privacy, and equity [14]. International organizations such as the World Health Organization (WHO) and the International Medical Device Regulators Forum (IMDRF) have issued updated guidelines emphasizing transparency, accountability, and inclusivity in the development and deployment of AI tools [14, 42]. WHO’s 2024 white paper on generative AI outlines key principles for managing large language models in healthcare, advocating for robust data stewardship and algorithmic explainability to prevent misuse and bias [42]. To bridge the gap between innovation and regulation, many high-income countries have launched regulatory sandboxes — controlled environments where AI developers, clinicians, and policymakers collaboratively test novel algorithms under real-world conditions while ensuring patient safety [43]. These sandboxes enable iterative refinement of AI models, facilitate faster regulatory approvals, and build public trust in AI-enabled clinical workflows. Equally critical is ensuring that ethical principles are not merely aspirational but are embedded in local policy and practice, especially in low- and middle-income countries where governance structures may be underdeveloped [14, 44]. Regional collaborations, standardized data sharing agreements, and cross-border ethical oversight mechanisms can help harmonize AI implementation across diverse healthcare systems (Fig. 3). Fig. 3Global comparison of digital health maturity levels. This figure was developed using publicly available data synthesized from WHO (2024 Digital Health Report), IMDRF (2022–2024 regulatory guidelines), and GA4GH (2021 genomic data harmonization initiatives). It compares readiness and maturity across high-income countries and low-/middle-income countries, with emphasis on implications for cardio-oncology Global comparison of digital health maturity levels. This figure was developed using publicly available data synthesized from WHO (2024 Digital Health Report), IMDRF (2022–2024 regulatory guidelines), and GA4GH (2021 genomic data harmonization initiatives). It compares readiness and maturity across high-income countries and low-/middle-income countries, with emphasis on implications for cardio-oncology Collectively, robust ethical governance and adaptive regulatory frameworks are foundational to ensuring that AI tools in cardio-oncology are safe, effective, and equitably accessible. Opportunities for equitable innovation Bridging the global digital divide in cardio-oncology requires not only robust governance but also innovative technological paradigms that accommodate diverse regional constraints. Two emerging approaches — federated learning and digital twins — offer promising pathways to equitable, high-impact AI deployment. Federated learning enables collaborative AI model training across multiple institutions without requiring raw patient data to leave local servers [45]. This technique preserves patient privacy and respects local data sovereignty while expanding access to diverse, representative datasets. In low- and middle-income countries (LMICs), where data sharing is often hampered by regulatory and infrastructure barriers, federated learning can facilitate the development of robust AI models tailored to local populations [45, 46]. Pilot studies in oncology and cardiac imaging have demonstrated the feasibility and accuracy of federated learning frameworks compared to traditional centralized training approaches [47]. A structured roadmap summarizing opportunities for equitable AI implementation is presented in Table 3. Table 3Roadmap to equitable AI implementation in cardio-oncology. This table was developed based on integration of global initiatives (GA4GH, WHO, IMDRF) and published expert recommendations. It outlines five focus areas—collaborative research, regulatory harmonization, education, infrastructure, and ethical practices—linking each to example initiatives and measurable metrics to guide future implementationFocus AreaKey RecommendationsExample InitiativesKey MetricsCollaborative Research- Build HIC-LMIC datasets- Tackle algorithm bias- GA4GH, WHO Digital Health- % LMIC data in AI modelsRegulatory Harmonization- Global frameworks- Assist LMIC policy drafting- IMDRF, ITU/UNESCO AI for Good- # LMICs with AI policiesEducation- Train LMIC clinicians- AI4Health, ME AI Academy- # trained per yearInfrastructure- Improve rural digital access- Vision 2030, Microsoft AI- Broadband coverage %Ethical Practices- Consent/privacy standardization- Transparent algorithms- GDPR, NSF/Amazon “Fairness”- Compliance audits Table 3Roadmap to equitable AI implementation in cardio-oncology. This table was developed based on integration of global initiatives (GA4GH, WHO, IMDRF) and published expert recommendations. It outlines five focus areas—collaborative research, regulatory harmonization, education, infrastructure, and ethical practices—linking each to example initiatives and measurable metrics to guide future implementationFocus AreaKey RecommendationsExample InitiativesKey MetricsCollaborative Research- Build HIC-LMIC datasets- Tackle algorithm bias- GA4GH, WHO Digital Health- % LMIC data in AI modelsRegulatory Harmonization- Global frameworks- Assist LMIC policy drafting- IMDRF, ITU/UNESCO AI for Good- # LMICs with AI policiesEducation- Train LMIC clinicians- AI4Health, ME AI Academy- # trained per yearInfrastructure- Improve rural digital access- Vision 2030, Microsoft AI- Broadband coverage %Ethical Practices- Consent/privacy standardization- Transparent algorithms- GDPR, NSF/Amazon “Fairness”- Compliance audits Roadmap to equitable AI implementation in cardio-oncology. This table was developed based on integration of global initiatives (GA4GH, WHO, IMDRF) and published expert recommendations. It outlines five focus areas—collaborative research, regulatory harmonization, education, infrastructure, and ethical practices—linking each to example initiatives and measurable metrics to guide future implementation - Build HIC-LMIC datasets - Tackle algorithm bias - Global frameworks - Assist LMIC policy drafting - Consent/privacy standardization - Transparent algorithms Digital twins, virtual replicas of individual patients built from real-time clinical, imaging, and genomic data, represent the next frontier for precision medicine in cardio-oncology [48]. By simulating treatment scenarios and predicting patient-specific responses, digital twins could optimize chemotherapy planning, cardioprotective interventions, and follow-up strategies, minimizing the risk of irreversible cardiac damage [47]. Although still in early research phases, advances in computational modeling and real-time data integration are accelerating the clinical translation of digital twins for complex cardiovascular conditions. Investing in these cutting-edge technologies, coupled with capacity-building and ethical governance, can help ensure that the benefits of AI-driven cardio-oncology care reach patients globally — not just those in resource-rich settings. Key challenges and roadmap ahead While artificial intelligence (AI) offers transformative potential for cardio-oncology, its equitable and sustainable integration faces persistent, real-world challenges that must be systematically addressed: Data bias and representativeness Many AI algorithms are developed using datasets that underrepresent minority and LMIC populations, risking biased predictions and health inequities [32, 48]. Developing diverse, region-specific data pipelines and investing in federated learning can mitigate this risk. Local validation and generalizability AI models trained in high-income settings may not generalize well to different genetic backgrounds, healthcare infrastructures, or treatment protocols [49]. Rigorous local validation and context-aware adaptation are essential before widespread clinical deployment. Workforce education and digital literacy Clinicians and allied health professionals need structured training to interpret AI outputs, understand model limitations, and maintain human oversight [49]. Collaborative academic–industry educational programs can accelerate this competency building. Sustainable funding and infrastructure Many LMICs lack reliable digital infrastructure and long-term funding to scale pilot AI projects into routine practice [50]. Policy incentives, cross-sector partnerships, and public–private funding models are crucial to overcome these barriers. Multidisciplinary stakeholders — including clinicians, policymakers, technologists, and patient advocates — must collaborate to create transparent, adaptable, and inclusive AI ecosystems. Only through collective commitment can AI’s promise be fully realized to advance equitable, high-quality cardio-oncology care worldwide. Fig. 4 provides a clear visual roadmap highlighting how collaborative research, targeted education, robust infrastructure, and ethical frameworks can together advance equitable AI implementation in cardio-oncology. Fig. 4Roadmap to equitable AI adoption in cardio-oncology. This figure visualizes the main challenges, enablers, and proposed strategies described in section "Key challenges and roadmap ahead". It provides a conceptual overview of how collaborative research, targeted education, digital infrastructure, and ethical governance can collectively drive equitable AI integration Roadmap to equitable AI adoption in cardio-oncology. This figure visualizes the main challenges, enablers, and proposed strategies described in section "Key challenges and roadmap ahead". It provides a conceptual overview of how collaborative research, targeted education, digital infrastructure, and ethical governance can collectively drive equitable AI integration While we have summarized key barriers and proposed strategies, a visual representation may further clarify the implementation pathway. To this end, we developed a new schematic (Fig. 5) that integrates challenges, enablers, and action steps into a unified roadmap. The figure highlights four domains—collaborative research, workforce education, infrastructure, and governance/ethics—each mapped to practical interventions that can accelerate AI adoption in cardio-oncology across diverse health systems. This complements Fig. 4 by focusing on a pragmatic, stepwise translation from barriers to solutions, emphasizing scalability and equity in both high-income and low-resource settings. Fig. 5Roadmap to AI–cardio-oncology integration: from challenges to outcomes. This schematic translates the narrative roadmap in section "Key challenges and roadmap ahead" into a visual flow. It maps four key challenges (data bias, workforce gaps, infrastructure/funding, regulatory/ethical fragmentation) to corresponding enablers (collaborations, training, digital infrastructure, governance) and expected outcomes (equitable adoption, early detection, workflow integration, improved survivorship) Roadmap to AI–cardio-oncology integration: from challenges to outcomes. This schematic translates the narrative roadmap in section "Key challenges and roadmap ahead" into a visual flow. It maps four key challenges (data bias, workforce gaps, infrastructure/funding, regulatory/ethical fragmentation) to corresponding enablers (collaborations, training, digital infrastructure, governance) and expected outcomes (equitable adoption, early detection, workflow integration, improved survivorship) Conclusion Artificial intelligence is redefining the landscape of cardio-oncology by enhancing early detection, refining risk stratification, enabling continuous monitoring, and supporting personalized clinical decisions. However, realizing its full potential demands urgent action to address data bias, infrastructure gaps, local validation, and ethical governance — especially in under-resourced settings. With coordinated investments, equitable innovation strategies, and sustained stakeholder collaboration, AI can truly bridge global care disparities and transform cardio-oncology into a more precise, patient-centered discipline for the years ahead. Supplementary Information Supplementary Material 1. Supplementary Material 1.