Artificial intelligence (AI) will revolutionise cardiology practices over the next decade, from optimising diagnostics to individualising treatment strategies. Moreover, it can play an important role in combating gender inequalities in cardiovascular disease outcomes. There is growing evidence that AI algorithms can match humans at echocardiography analysis, while also being able to extract subtle differences that the human eye cannot detect. Similar promise is evident in the analysis of electrocardiograms, creating a new layer of interpretation. From big data, AI can produce algorithms that individualise cardiac risk factors and prevent perpetuating gender biases in diagnosis. Nonetheless, AI implementation requires caution. To avoid worsening health inequalities, it must be trained across diverse populations, and when errors arise, a robust regulatory framework must be in place to ensure safety and accountability. AI is perfectly positioned to capitalise on the growth of big data, but to proceed we require a generation of physicians who understand its fundamentals.
Introduction
Artificial intelligence (AI) is poised to revolutionise cardiology over the next decade, offering unprecedented potential and exciting advancements. The immense burden of cardiovascular disease in the population provides cardiologists with a huge swathe of rich medical data, yet at the moment this is still underutilised. Machine learning and deep learning are subsets of AI that learn from data, rather than being specifically programmed, to identify new patterns and produce decision-making models.1 From improving diagnostic accuracy to enhancing treatment strategies, machine learning has the power to reshape patient care and outcomes in ways that have yet to be fully grasped. Moreover, AI can play a crucial role in addressing gender inequalities in cardiovascular disease (CVD) outcomes. However, as with any new technology, it is crucial that any potential challenges and dilemmas are addressed prior to implementation. Here, I will explore these arguments and highlight the ways in which AI will change cardiology, with a focus on advanced diagnostic imaging, risk prediction and decision-making.
Imaging
One key area where AI can bring significant advancements is in diagnostic imaging. Machine-learning algorithms can analyse images from echocardiography, cardiac magnetic resonance imaging (MRI), and computed tomography (CT) angiography, to detect subtle differences and extract hidden signals that the human eye cannot see.2 Within echocardiography, these algorithms have been shown to be as good as humans at segmenting the cardiac chambers and calculating simple measurements across the five common views,3 with an added benefit of reduced inter- and intra-operator variability.3 Overall, with a large enough sample size to train on, AI-driven imaging tools can standardise measurements, reduce human errors and assist cardiologists in making more precise and timely diagnoses.
Even the humble electrocardiogram (ECG) can now be greatly enhanced by AI. Deep-learning algorithms are now known to be able to match, or even outperform, cardiologists in recognising arrhythmias.4 Additionally, a recent pragmatic trial has demonstrated that AI-enabled ECGs can be used as a low-cost non-invasive screening tool, by identifying asymptomatic patients with a high likelihood of a reduced ejection fraction.5,6 ECG changes were detected even before echocardiography evidence, with a positive AI algorithm result representing a fourfold higher future risk of a low ejection fraction.6 By utilising AI to identify patients at risk of developing symptomatic heart failure, guideline-directed medical therapy can be initiated early, ultimately improving long-term patient outcomes and reducing healthcare-associated costs.
Individualisation
Another exciting prospect of AI use in cardiology lies in personalised medicine. By integrating patient-specific data, such as genetics, biomarkers, and clinical records, AI algorithms could help tailor treatment plans to individual patients. For instance, AI algorithms have already shown effectiveness in predicting individual responses to cardiac resynchronisation therapy (CRT) before implantation.7,8 This is an important step, as implantation is expensive and invasive, yet around a third experience little clinical benefit from CRT.9 Traditionally, treatment decisions are based on large trial data, which while effective at the population level, are less so for an individual. Using machine learning, cardiologists will be better able to predict an individual patient’s clinical outcomes, thereby improving the shared decision-making process and reducing unnecessary interventions.
Equality
AI can take advantage of big data and transform cardiovascular medicine. However, there is concern that the historic underrepresentation of women in cardiovascular research risks perpetuating gender biases and exacerbating disparities in diagnosis, treatment, and outcomes.10 Women have unique sex-specific cardiovascular risk factors, additional to those traditionally associated with cardiovascular disease (CVD), which can influence their CVD burden, and lead to atypical presentations and misdiagnosis.11 Nevertheless, with the right direction, AI presents the opportunity to address these gaps and deliver sex-specific diagnostic tools.11 One such example is the identification of acute myocardial infarction (MI). Despite troponin levels varying with age and sex,12 the cut-offs typically applied are universal and, thus, may miss true cases of MI in women. Recently, a machine-learning algorithm has shown promise in calculating individualised MI probabilities, based on clinical factors and troponin levels, with equitable results across sexes and ages.13 By employing AI-driven decision tools trained on diverse datasets, we can gain deeper insights into how cardiovascular diseases affect different patient groups and develop more effective strategies for prevention, diagnosis, treatment, and medication selection. This has the potential to reduce gender-based disparities in cardiovascular outcomes, marking an exciting advancement towards more equitable healthcare.
Reasons for concern
Therefore, while AI holds great potential in cardiology, its implementation should be approached with caution, considering ethics, data bias, model overfitting and the need for human oversight. AI models rely entirely upon the information they are trained on and, thus, have the potential to propagate gender and racial biases, as discussed earlier.14 To mitigate this, and ensure their effectiveness and generalisability, AI models must be trained and validated across diverse populations using representative data.
One of the other significant concerns raised by AI is similar to the debates surrounding driverless cars: who is accountable for errors that may arise when using AI? If a clinician disagrees with an AI output, who bears the responsibility for any negative outcomes? These questions require careful consideration prior to the widespread deployment of AI tools, with the answers varying depending on the nature of the tool. Models designed to inform clinical decision-making may be treated akin to laboratory or radiological tests regarding licensing. On the other hand, models that directly interact with patients may require more rigorous scrutiny if their application is not filtered through a clinician. The establishment of robust regulatory frameworks is vital to ensure the safety, efficacy, and accountability of AI in cardiology. By addressing these challenges today, we can maximise the potential benefits of AI in cardiology, while safeguarding against potential pitfalls tomorrow.
Conclusion
In conclusion, the integration of AI into cardiology holds immense promise over the coming decade. Cardiology, with its vast datasets and highly evidence-based guidelines, is perfectly positioned to capitalise. The technology is available, but the existing workforce knowledge base and medico-legal infrastructure are real-time barriers to widespread implementation. This requires improving clinicians’ understanding of AI, beginning at medical school, with the introduction of algorithmics and computer science into the curriculum. By fostering collaboration among AI researchers, regulators, and policymakers, cardiologists can establish AI as an indispensable tool to enhance patient outcomes, promote equitable care, and shape the future of cardiology.
Key messages
- Artificial intelligence will revolutionise cardiology practices over the next decade, including imaging and risk factor identification
- Artificial intelligence can potentially help address gender differences in cardiovascular disease outcomes
- Artificial intelligence will need to be taught at medical school and requires robust regulatory frameworks before its widespread implementation
Conflicts of interest
None declared.
Funding
None.
Editors’ note
This article was the prize-winning essay in the National Essay Prize 2023 of the British Junior Cardiologists’ Association (BJCA).
References
1. Karatzia L, Aung N, Aksentijevic D. Artificial intelligence in cardiology: hope for the future and power for the present. Front Cardiovasc Med 2022;9:945726. https://doi.org/10.3389/fcvm.2022.945726
2. Lin A, Kolossváry M, Motwani M et al. Artificial intelligence in cardiovascular imaging for risk stratification in coronary artery disease. Radiol Cardiothorac Imaging 2021;3:e200512. https://doi.org/10.1148/ryct.2021200512
3. Zhang J, Gajjala S, Agrawal P et al. Fully automated echocardiogram interpretation in clinical practice. Circulation 2018;138:1623–35. https://doi.org/10.1161/CIRCULATIONAHA.118.034338
4. Hannun AY, Rajpurkar P, Haghpanahi M et al. Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network. Nat Med 2019;25:65–9. https://doi.org/10.1038/s41591-018-0268-3
5. Yao X, Rushlow DR, Inselman JW et al. Artificial intelligence-enabled electrocardiograms for identification of patients with low ejection fraction: a pragmatic, randomized clinical trial. Nat Med 2021;27:815–19. https://doi.org/10.1038/s41591-021-01335-4
6. Attia ZI, Kapa S, Lopez-Jimenez F et al. Screening for cardiac contractile dysfunction using an artificial intelligence-enabled electrocardiogram. Nat Med 2019;25:70–4. https://doi.org/10.1038/s41591-018-0240-2
7. Tokodi M, Schwertner WR, Kovács A et al. Machine learning-based mortality prediction of patients undergoing cardiac resynchronization therapy: the SEMMELWEIS-CRT score. Eur Heart J 2020;41:1747–56. https://doi.org/10.1093/eurheartj/ehz902
8. Kalscheur MM, Kipp RT, Tattersall MC et al. Machine learning algorithm predicts cardiac resynchronization therapy outcomes: lessons from the COMPANION trial. Circ Arrhythm Electrophysiol 2018;11:e005499. https://doi.org/10.1161/CIRCEP.117.005499
9. McAlister FA, Ezekowitz J, Hooton N et al. Cardiac resynchronization therapy for patients with left ventricular systolic dysfunction: a systematic review. JAMA 2007;297:2502–14. https://doi.org/10.1001/jama.297.22.2502
10. Vogel B, Acevedo M, Appelman Y et al. The Lancet women and cardiovascular disease commission: reducing the global burden by 2030. Lancet 2021;397:2385–438. https://doi.org/10.1016/S0140-6736(21)00684-X
11. Adedinsewo DA, Pollak AW, Phillips SD et al. Cardiovascular disease screening in women: leveraging artificial intelligence and digital tools. Circ Res 2022;130:673–90. https://doi.org/10.1161/CIRCRESAHA.121.319876
12. de Bakker M, Anand A, Shipley M et al. Sex differences in cardiac troponin trajectories over the life course. Circulation 2023;147:1798–808. https://doi.org/10.1161/CIRCULATIONAHA.123.064386
13. Doudesis D, Lee KK, Boeddinghaus J et al. Machine learning for diagnosis of myocardial infarction using cardiac troponin concentrations. Nat Med 2023;29:1201–10. https://doi.org/10.1038/s41591-023-02325-4
14. Paulus JK, Kent DM. Predictably unequal: understanding and addressing concerns that algorithmic clinical prediction may increase health disparities. NPJ Digit Med 2020;3:99. https://doi.org/10.1038/s41746-020-0304-9
