Abbreviations: AI, Artificial intelligence; CVD, Cardiovascular disease; FEMTECH, Female technology; ML, Machine learning.
Introduction
Cardiovascular disease (CVD) is the leading cause of death among women in the United States and worldwide [1], including young women [2]. Unfortunately, CVD in women is frequently “understudied, underdiagnosed, and undertreated” [3, 4]. As such, cardiovascular health promotion, early detection, and management of CVD are essential to decrease the risk of death and disability. With recent advancements in technology and the advent of personalized digital health devices, artificial intelligence (AI) has emerged as a promising solution for improving cardiovascular health among women [5–8].
In this Commentary article, we provide a succinct overview of AI solutions to improve women’s cardiovascular health, and highlight the potential pitfalls of AI and novel digital technologies.
Artificial Intelligence Solutions for Improving Cardiovascular Health among Women
AI is a form of machine learning (ML) that broadly describes the ability of machines to perform intelligent tasks, such as learning unique features from large datasets and making predictions based on available data. The ability of AI to synthesize large amounts of data in short periods of time far exceeds human capabilities. Thus, AI serves as a unique tool to leverage the vast amounts of digital data generated in the past few years from electronic medical record documentation, diagnostic imaging tests, personalized biometric monitoring technologies/wearables, and even social media. Published studies are increasingly reporting the use of ML in cardiovascular health care [9]. However, few studies have focused on the use of AI tools specifically among women. A few use cases are discussed below.
AI can be used to provide personalized cardiovascular care to women through cardiovascular health promotion (Figure 1). Notably, AI-powered mobile health apps with guided feedback and participatory elements, such as food scanners, quizzes, and games, have been used as interactive educational and practical decision-support tools enabling patients to engage in self-care, and encouraging the adoption of ideal cardiovascular health behaviors [10–12].
These novel technologies can also be leveraged to improve the detection and management of CVD among women. In a recent article, we discussed the immense potential of AI and digital tools to improve CVD screening across a woman’s life course [13]. Specific AI studies targeted at early disease detection in women include the identification of left ventricular dysfunction during pregnancy and the postpartum period [14, 15]; the use of mammography for automated assessment of breast arterial calcification, a known risk factor for CVD [16]; and early detection of gestational diabetes, preeclampsia, and other pregnancy complications [17]. However, many of these studies lack rigorous external validation and have yet to be evaluated prospectively or in clinical trials. AI has also shown some utility in interpreting cardiovascular test data, such as those from echocardiography, computed tomography, and cardiac magnetic resonance imaging. AI algorithms have been shown to have potential roles in improving the diagnosis of cardiovascular disorders such as heart failure, myocardial infarction, arrhythmias, and valvular heart disease [18].
There is a need to improve current cardiovascular risk prediction tools as many existing tools do not incorporate female-specific risk factors, such as adverse pregnancy outcomes, polycystic ovarian syndrome, and menopause [1, 19]. Some studies have shown that ML algorithms can improve the prediction of risk and health outcomes among women by identifying and incorporating novel risk factors for atherosclerotic CVD [20], incident heart failure [21], and mortality [22].
The use of digital technologies can also help increase the participation of women in clinical trials [23], particularly decentralized and digitally enabled studies, thus narrowing the sex and gender gap in CVD research and outcomes. A few digitally enabled cardiovascular trials have demonstrated effectiveness in enrolling women [24, 25]. The potential roles of AI and digital technologies in women’s health are increasingly being recognized, and the development of specific technologies in this field has been broadly described as “femtech” [26]. Although these technologies have traditionally focused on reproductive health, many are starting to explore the cardiovascular health space [27, 28].
Artificial Intelligence Challenges and Pitfalls
Although advances in digital technologies and AI have immense potential to improve women’s cardiovascular health, the potential risk of harm that may result from the rapid adoption of these technologies cannot be ignored. Several cardiovascular clinical trials and studies largely include men [3], thus resulting in potential bias if AI models are developed using these datasets. Multiple studies have demonstrated the potential for AI to perpetuate and reinforce inequities – termed algorithmic bias [29]. The Gender Shades study identified a major failure of commercial facial recognition algorithms to correctly identify women with darker skin tones while showing near-perfect performance among White men [30]. Other studies have also demonstrated race and sex bias in AI algorithms intended for use in healthcare [31, 32]. Therefore, it is important that AI-based tools in cardiovascular medicine utilize diverse datasets during development, are rigorously tested, broadly validated, and have safeguards in place to limit potential harm from its use. These efforts are referred to as responsible AI [33].
Prior to broad implementation of AI tools, it is important that we consider the potential for these technologies to be scaled across healthcare systems, clinical/non-clinical settings, and worldwide. Portable, efficient, and inexpensive digital devices are necessary to deploy AI models, including considerations for edge computing vs. cloud-based computing capabilities. In addition, AI-based solutions must be capable of integration with existing systems to maximize the benefits of AI investments. Interoperability issues can arise from a lack of standardization in data formats, which can result in difficulties transferring data between systems. AI scalability and interoperability challenges can be addressed by using open-source tools in addition to developing and implementing standards that facilitate data exchange, such as blockchain technology [34].
Also important is the cost implication of adopting novel AI and digital technologies. These include the cost of setting up the infrastructure necessary for deploying AI tools; purchasing and maintaining hardware and software; and hiring AI experts and/or training personnel to use AI. Furthermore, the data used to train AI models must be collected, stored, and analyzed securely. Some studies have evaluated the cost associated with implementing specific AI tools [35–37], and demonstrated their cost-effectiveness and potential to improve health outcomes.
Finally, digital inclusion and equity must be considered as essential social determinants of health that could influence the adoption and impact of AI and digital technologies among women [38]. For all women to benefit from AI-driven solutions to improve cardiovascular health, ensuring access to digital technologies and affordability are key to narrowing the digital divide [39]. While a digital gap is apparent in the United States, the gap between high-income countries and low- or middle-income countries is even wider, due to cost and infrastructure constraints [40]. However, these challenges are not insurmountable. Evidence indicates that low- and middle-income countries can use and benefit from novel technologies [41], but a commitment towards ensuring digital equity is required from all stakeholders, including the technology industry, government and policymakers, health researchers, and healthcare institutions [13].
Conclusion
The use of AI to improve women’s cardiovascular health has great promise; however, we must be aware of potential pitfalls, be intentional in developing and implementing strategies to address bias, and guard against the misuse of technologies that could cause harm. It is essential that novel digital technologies and AI tools intended for use among women are thoroughly evaluated in prospective studies, clinical trials, diverse populations, and in implementation studies – to assess the feasibility of incorporating these tools into existing healthcare models. We believe that the promise of AI-driven solutions can be achieved with targeted investments and commitment from all stakeholders in the field of cardiovascular health.