AI: A catalyst or inhibitor for health equity in Radiology?

The webinar explored the dual nature of AI in the context of healthcare equity. Following panelists discussed how AI can act as both a catalyst and an inhibitor of healthcare equity, providing practical strategies to address disparities in medical imaging and healthcare as a whole. To view the full webinar visit https://global.medical.canon/News/Events/perspectives-webinar-20240903.html

Health Equity in Radiology: The Current Landscape

The World Health Organization defines health equity as the absence of unfair and avoidable health differences among population groups based on social, economic, demographic, or geographic factors, which impacts the distribution of resources, and opportunities.1 Despite this, healthcare disparities persist across all medical specialties, including radiology with two-thirds of the global population lacking adequate access to radiology services.2,3

Patient level barriers such as access to reliable transportation, insurance, educational level and financial constraints contribute to these disparities.2,4 In low- and middle-income countries (LMICs), disparities are worsened by shortage of radiologists, inadequate equipment and poor basic infrastructure.5 The International Atomic Energy Agency reveals a staggering disparity factor of over 65-fold, with high-income countries having one CT scanner per 25,000 people, while low-income nations have one per 1.69 million people. (Figure 1).3

The current radiology system is data-rich but information-poor, meaning that despite the abundance of radiology data, it is not being fully utilized to directly benefit patient care.6 The goal is to progress from data to knowledge and, ultimately, to wisdom—an achievement that can be facilitated by artificial intelligence (AI).

While the use of AI in radiology has grown exponentially in recent years (Figure 2), its inequitable application has been shown to widen healthcare disparities in some cases.3,7 Additionally, inadequate information technology (IT) infrastructure in LMICs poses a significant challenge to AI implementation.3

The current radiology system is data-rich but information-poor, meaning that despite the abundance of radiology data, it is not being fully utilized to directly benefit patient care.6 The goal is to progress from data to knowledge and, ultimately, to wisdom—an achievement that can be facilitated by artificial intelligence (AI).

While the use of AI in radiology has grown exponentially in recent years (Figure 2), its inequitable application has been shown to widen healthcare disparities in some cases.3,7 Additionally, inadequate information technology (IT) infrastructure in LMICs poses a significant challenge to AI implementation.3

Hence, there is an urgent need to explore how we can improve the implementation of AI to bridge these disparities.

Leveraging AI to Democratize Medical Imaging: Bridging the Gap Between 'Haves' ?and 'Have-Nots’

Improving access and availability

The integration of AI, radiology and telemedicine has the potential to transform remote diagnostics and patient care by overcoming geographical barriers, expanding access to specialized expertise, and improving diagnostic efficiency and accuracy. Telemedicine enables patients to securely transmit medical images, allowing radiologists to review them from any location. AI-driven image analysis further enhances diagnoses by delivering prompt and accurate results, even in areas with limited radiological expertise. This advancement improves access to radiology services and ensures quality care, regardless of location.8

Additionally, the emergence of AI tools, capable of enhancing efficiency and efficacy at every stage—from scheduling to follow-up monitoring—has empowered radiology departments to implement new strategies and workflows that can help mitigate workforce shortages.9,10 AI-based virtual assistants may also help address the chronic radiologist shortage by alleviating the tedious task of dictating long radiology reports.11 Furthermore, AI optimizes personnel allocation, enhances scanner usage, and minimizes radiation exposure, thereby improving both efficiency and quality of care.10

RAD-AID, a nonprofit organization, has devised a three-pronged strategy for low-resource health institutions, focusing on clinical radiology education, infrastructure implementation, and a phased introduction of AI. The goal is to significantly enhance access to radiology services through the deployment of picture archiving and communication system (PACS) software and the integration of radiology hardware.3

Reducing Bias and Fostering Trust

Although AI has the potential to enhance health equity,? it also carries the risk of introducing bias.10 The main sources of bias in healthcare AI include data, algorithm, AI-clinician interactions and AI-patient interactions.12 To mitigate these biases, strategies consist of leveraging diverse datasets,12,13 conducting regular audits, and engaging external experts to accurately identify and eliminate biases.12

Developing ethical and responsible AI is crucial for building and maintaining trust among clinicians, patients, and the public, which facilitates its implementation in clinical practice. Robust governance throughout all stages of the AI lifecycle requires the involvement of key stakeholders, including software developers, government bodies, clinical practitioners, and patient interest groups. Figure 3 summarizes the principles related to ethical and responsible AI for radiology and radiography in Europe.14

The European Union AI Act, effective August 2024, is the world’s first comprehensive legal framework for AI, designed to foster innovation while safeguarding individuals from potential harm. The AI Act adopts a risk-based approach: the higher the risk, the stricter the regulations. Health-related AI systems are categorized into different risk levels, with applicable requirements based on whether the system is classified as high-risk, low-risk, minimal-risk, or a general-purpose AI model.15

High-risk AI systems, such as diagnostic tools, must meet safety and quality standards, including third-party assessments.15 Approximately 75% of all commercial AI-enabled medical devices are related to radiology, with most classified as Class ≥IIa under the medical device regulations (MDR). As high-risk devices, they must comply with the AI Act's requirements within 12 months of its implementation.16

As healthcare continues to drive AI advancements, the new regulations will transform national approaches to health technology policies and practices.15

Enhancing AI education in radiology:

To support AI’s ongoing growth in radiology, radiologists must be adept at using AI tools and influencing the creation of clinically useful applications. Foundational courses, such as the National Imaging Informatics Course – Radiology (NIIC-RAD) and the Radiological Society of North America (RSNA) Imaging AI Certificate, provide trainees with the necessary framework to explore the creation, deployment, and evaluation of AI applications. The NIIC-RAD course provides an introduction to imaging informatics, with a hybrid format for self-guided learning and expert discussions. Meanwhile, the RSNA Imaging AI Certificate offers structured education, including on-demand lectures and hands-on exercises for evaluating and deploying AI tools. These AI education initiatives will set important precedents for radiology training, empowering future generations to adapt to evolving paradigms in imaging AI. 17

Summary:

AI in healthcare, especially in medical imaging, is rapidly advancing.7 However, rising social and economic barriers are widening health inequities. 2,4 AI holds significant potential to promote health equity in radiology by improving access, reducing bias and enhancing education.3,18?The successful integration of AI depends upon three critical factors: clinical radiology education, infrastructure implementation, and phased AI introduction.3 While AI presents numerous opportunities, it also poses ?challenges.3,12 Developing ethical and responsible AI is crucial for maintaining trust among clinicians, patients, and the public, ensuring? its implementation in clinical practice.14 When used wisely, AI can be a catalyst for health equity in radiology.

This commentary is the product of round table discussions convened and facilitated by Canon Medical Systems Corporation, held virtually on Sep 3, 2024. To view the full webinar visit https://global.medical.canon/News/Events/perspectives-webinar-20240903.html

References:

1.?????? Health equity WPRO. Available from: https://www.who.int/westernpacific/health-topics/equity#tab=tab_1 Accessed on 09 December 2024.

2.?????? DeBenedectis CM, Spalluto LB, Americo L, et al. Health Care Disparities in Radiology-A Review of the Current Literature.?J Am Coll Radiol. 2022;19(1 Pt B):101-111.

3.?????? Mollura DJ, Culp MP, Pollack E, et al. Artificial Intelligence in Low- and Middle-Income Countries: Innovating Global Health Radiology.?Radiology. 2020;297(3):513-520.

4.?????? Spalluto LB, Friedman E, Sonubi C, et al. Equality Is Not Fair: Imaging and Imagining the Road to Health Equity.?J Am Coll Radiol. 2022;19(1 Pt B):139-142.

5.?????? Hinrichs-Krapels S, Tombo L, Boulding H, et al. Barriers and facilitators for the provision of radiology services in Zimbabwe: A qualitative study based on staff experiences and observations.?PLOS Glob Public Health. 2023;3(4):e0001796.

6.?????? Kharat AT, Singh A, Kulkarni VM, et al. Data mining in radiology.?Indian J Radiol Imaging. 2014;24(2):97-102.

7.?????? Liu DS, Abu-Shaban K, Halabi SS, et al. Changes in Radiology Due to Artificial Intelligence That Can Attract Medical Students to the Specialty.?JMIR Med Educ. 2023;9:e43415.

8.?????? Bo?i? V. Radiology, Telemedicine and Artifical Intelligence. 10.13140/RG.2.2.34259.96800.

9.?????? Kalidindi S, Gandhi S. Workforce Crisis in Radiology in the UK and the Strategies to Deal With It: Is Artificial Intelligence the Saviour?.?Cureus. 2023;15(8):e43866.

10.?? Najjar R. Redefining Radiology: A Review of Artificial Intelligence Integration in Medical Imaging.?Diagnostics (Basel). 2023;13(17):2760.

11.?? Langlotz CP. The Future of AI and Informatics in Radiology: 10 Predictions.?Radiology. 2023;309(1):e231114.

12.?? Ueda D, Kakinuma T, Fujita S, et al. Fairness of artificial intelligence in healthcare: review and recommendations.?Jpn J Radiol. 2024;42(1):3-15.

13.?? Sitek A. Artificial Intelligence in Radiology: Bridging Global Health Care Gaps through Innovation and Inclusion.?Radiol Artif Intell. 2024;6(2):e240093.

14.?? Walsh G, Stogiannos N, van de Venter R, et al. Responsible AI practice and AI education are central to AI implementation: a rapid review for all medical imaging professionals in Europe.?BJR Open. 2023;5(1):20230033.

15.?? van Kolfschooten H, van Oirschot J. The EU Artificial Intelligence Act (2024): Implications for healthcare.?Health Policy. 2024;149:105152.

16.?? Busch F, Kather JN, Johner C, et al. Navigating the European Union Artificial Intelligence Act for Healthcare.?NPJ Digit Med. 2024;7(1):210.

17.?? Tejani AS, Elhalawani H, Moy L, et al. Artificial Intelligence and Radiology Education.?Radiol Artif Intell. 2022;5(1):e220084.

18.?? Khalifa M, Albadawy M. AI in diagnostic imaging: Revolutionising accuracy and efficiency. Computer Methods and Programs in Biomedicine Update. 2024:100146. DOI:10.1016/j.cmpbup.2024.100146