Interventional Cardiologists’ Perspectives and Knowledge Towards Artificial Intelligence

© 2024 HMP Global. All Rights Reserved.
Any views and opinions expressed are those of the author(s) and/or participants and do not necessarily reflect the views, policy, or position of the Journal of Invasive Cardiology or HMP Global, their employees, and affiliates. 

Watch the accompanying author interview here.

Background. Artificial intelligence (AI) is increasingly utilized in interventional cardiology (IC) and holds the potential to revolutionize the field.

Methods. We conducted a global, web-based, anonymous survey of IC fellows and attendings to assess the knowledge and perceptions of interventional cardiologists regarding AI use in IC.

Results. A total of 521 interventional cardiologists participated in the survey. The median age range of participants was 36 to 45 years, most (51.5%) practice in the United States, and 7.5% were women. Most (84.7%) could explain well or somehow knew what AI is about, and 63.7% were optimistic/very optimistic about AI in IC. However, 73.5% believed that physicians know too little about AI to use it on patients and most (46.1%) agreed that training will be necessary. Only 22.1% were currently implementing AI in their personal clinical practice, while 60.6% estimated implementation of AI in their practice during the next 5 years. Most agreed that AI will increase diagnostic efficiency, diagnostic accuracy, treatment selection, and healthcare expenditure, and decrease medical errors. The most tried AI-powered tools were image analysis (57.3%), ECG analysis (61.7%), and AI-powered algorithms (45.9%). Interventional cardiologists practicing in academic hospitals were more likely to have AI tools currently implemented in their clinical practice and to use them, women had a higher likelihood of expressing concerns regarding AI, and younger interventional cardiologists were more optimistic about AI integration in IC.

Conclusions. Our survey suggests a positive attitude of interventional cardiologists regarding AI implementation in the field of IC.
 

Artificial intelligence (AI) is increasingly utilized in interventional cardiology (IC) and holds the potential to revolutionize the field.^1,2 We sought to systematically evaluate physicians’ attitudes related to the implementation of AI in clinical practice. We conducted a global, web-based survey of IC fellows and attendings to assess the knowledge and perceptions of interventional cardiologists regarding AI use in cardiology and specifically in IC.

The survey questionnaire was collaboratively developed by the co-authors through an interactive process. The final survey consisted of 35 questions in the English language and encompassed 3 domains: (1) demographic information; (2) perspectives on AI in cardiology; and (3) perspectives on AI tools in IC (Appendix). The survey was conducted using REDCap (Research Electronic Data Capture)^3,4 and was distributed to IC attendings and fellows via social media and email lists (6672 emails). The study was approved by the institutional review board of Minneapolis Heart Institute.

Categorical variables were expressed as percentages and compared using the Pearson’s chi-square test. Continuous variables are presented as mean ± standard deviation or as median (interquartile range) unless otherwise specified and were compared using the independent-samples t-test for normally distributed variables and the Mann-Whitney U test for non-parametric variables, as appropriate. All statistical analyses were performed using R Statistical Software, version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria). A P-value of less than .05 was considered statistically significant.

A total of 521 ICs (429 attendings and 92 fellows) participated in the survey. The median age range of the participants was 36 to 45 years. Among respondents, 7.5% were women and most participants practiced in the United States (51.5%), followed by the European Union (15.6%). Most participants practiced in academic settings (university-based, 31.7%; university-affiliated, 17.1%).

Among survey participants, 46.4% somewhat knew what AI is about, while 38.5% could explain it well. High/very high level of enthusiasm about AI in cardiology was expressed by 77.2% of the participants, while 63.7% were optimistic/very optimistic about AI in IC. Most participants believed that the field that will benefit the most from AI is advanced cardiac imaging (56.8%), followed by IC (13.3%), general cardiology (11.0%), and preventive cardiology (10.3%). Most participants anticipated implementation of AI in their personal clinical practice within 5 years (60.6%), while 15.7% reported that AI is currently implemented in their practice, and 7.4% may have access to AI in their clinical practice but they were not using it at the time of the survey

Most participants believed that AI would increase diagnostic efficiency, diagnostic accuracy, treatment selection, patient satisfaction, and healthcare expenditure, and decrease medical errors and physician workload (Figure 1). However, they expressed concerns, mainly regarding the limited external validity of AI tools, the AI errors, the low quality of data, the unclear framework, the cognitive bias of AI, and the accessibility to AI software (Supplemental Figure).

The most tried AI-powered tools were image analysis (57.3%), electrocardiogram (ECG) analysis (61.7%), and AI-powered algorithms (45.9%). These were also the tools that respondents believed have the higher potential for implementation in the field (Figure 2).

Most participants (76.8%) were not concerned about displacement from AI, and 46.3% believed that AI will not impact interventional cardiologists’ compensation (Figure 3).

Most participants (87.0%) would override AI based on their experience and knowledge and believed that the physician should always have the final control over diagnosis and therapy (94.4%). Regarding AI-related errors, most agreed that the physician making the decision should be primarily responsible (44.7%) or responsibility should be equally shared by multiple parties (35.5%) (Figure 3). However, a significant percentage of interventional cardiologists agreed that physicians know too little about AI to use it on patients (73.5%) and training will be necessary (64.8%) (Figure 3). They also expressed worries that AI-based systems could be manipulated from the outside (95.8%), prevent physicians from learning to make their own judgment of the patient (53.5%), and that physicians are becoming too dependent on computer systems (60.8%).

Younger interventional cardiologists (≤ 45 years) were more likely to consider themselves to have good command of digital technologies (82.3% vs 67.1%; P < .001) and to be very optimistic about AI implementation in IC. They more often believed that AI could increase diagnostic efficiency and patient satisfaction, decrease workload, and positively impact interventional cardiologists’ compensation (Figure 4A).

Women interventional cardiologists more often believed that patients will be negative/suspicious about AI integration in their care (52.9% vs 25.3%; P = .002) and more often expressed concern about the limited external validity of AI, real-time applicability, invasion of patients’ privacy, unknown/unclear regulatory framework, cognitive bias of AI, AI errors, lack of empathy, and compassion and accessibility to AI software (Figure 4B).

Interventional cardiologists with a good command of digital technologies were more likely to be very enthusiastic about AI in cardiology and very optimistic about AI in IC specifically, while they were more often able to explain well what AI is about or consider themselves experts in the field. They were also more likely to have used AI-enabled electrocardiogram analysis and AI-enabled virtual assistants, compared with interventional cardiologists who did not identify as having a good command of digital technologies (Figure 4C).

Participants practicing in academic hospitals were more likely to be able to explain well what AI is about (41.4% vs 35.7%; P = .007) or consider themselves experts in the field of AI (7.6% vs 2.3%; P = .007). They were more likely to currently use AI tools in their clinical practice (20.6% vs 11.0%; P = .015) and to express a high level of concern for the limited external validity of AI tools (51.1% vs 39.2%; P = .039) (Figure 4D).

Our study indicates that most interventional cardiologists have a positive attitude regarding AI. The main findings of our study are that (1) even though most interventional cardiologists anticipate implementation of AI in their clinical practice within the next 5 years, most agree that physicians currently know too little about AI to use it on patients; (2) most would override AI and agree that the physician making the decision should be primarily responsible for AI-related errors; (3) interventional cardiologists practicing in academic hospitals are more likely to have AI tools currently implemented in their clinical practice and to use them; (4) women have a higher likelihood of expressing concerns regarding AI; and (5) younger interventional cardiologists are more optimistic about AI integration in IC.

There is a demand for AI-related training to facilitate safe adoption, which should commence in the early stages of medical education and continue as part of the postgraduate training. Interventional cardiologists need to be knowledgeable about both the capabilities and limitations of AI, enabling them to interpret results accurately and navigate ethical challenges.⁵ Additionally, addressing concerns regarding the interpretability and explainability of AI tools and ensuring data quality and external validity should be a priority in the move towards transparent AI.⁶

Concerns also arise regarding legal liability, and these are expected to escalate with the growing complexity of AI advancements. Discussions about medical malpractice involving AI-related decision-making should be thorough and address potential liability gaps, while physicians need to be well-informed about their rights and personal liability when using AI tools in clinical practice. The prevailing sentiment among the majority of participants is that blind reliance on AI tools is inappropriate and physicians should bear the ultimate responsibility and be capable of overriding the AI. This sentiment aligns with current laws, which treat AI as a confirmatory tool rather than a primary decision-making tool.⁷ Concerns about job displacement or negative impact on compensation are limited, which aligns with the understanding that work in IC is far from being fully automated.²

However, it is essential to be aware that AI can outperform guideline-recommended pathways, demonstrating intelligence that can be too significant to be overlooked.⁸ Further, powerful large language models, such as ChatGPT (OpenAI) and other chatbots, have great potential to reduce documentation burden and improve workflows in daily clinical practice and even boost research productivity.⁹ Yet, the spectrum of capabilities and potential weaknesses of these otherwise impressive and ever-improving models is increasingly recognized.¹⁰ Education and increased vigilance among clinicians about the best practices for using such models are needed. 

There is consensus on the perspectives of survey participants regarding the positive impact of AI on clinical care. Nevertheless, variations in perspectives suggest that individuals older than 45 years are less optimistic about the implementation of AI in clinical care. On the other hand, women express more concerns about the potential risks that merit consideration. Academic physicians involved in clinical care are more likely to possess deeper knowledge about AI. This likely reflects their greater access to AI tools stemming from the more rapid adoption and implementation of these tools in academic healthcare settings.

Limitations. The limitations of the study include the possibility that participating IC attendings and IC fellows might have a higher interest in AI and are more likely to have stronger opinions on the topic, which might result in selection bias. Distribution by email lists and social media (X) could also introduce selection bias, potentially favoring younger and academic interventional cardiologists. Additionally, the study's findings are based on the current attitudes of participants, and future developments in AI could impact the relevance of the results over time. The present study did not examine specific AI tools currently used in interventional cardiology, such as FFRangio (CathWorks).

Our study highlights the strong interest regarding AI and growing demand for AI-related training among interventional cardiologists. Over the next 5 years, most of the participants anticipate the integration of AI into their clinical practice. Notably, those in academic hospitals are more likely to have AI tools currently in use, and younger interventional cardiologists and those with digital technology proficiency express higher optimism about AI in IC. The optimal ways to prepare the current and future IC workforce for the safe and effective integration of the surging AI technology in clinical practice are yet to be determined.

From the ¹Minneapolis Heart Institute and Minneapolis Heart Institute Foundation, Abbott Northwestern Hospital, Minneapolis, Minnesota, USA; ²Athens Naval Hospital, Athens, Greece; ³Center for Digital Cardiovascular Innovations, Division of Cardiovascular Medicine, University of Miami Miller School of Medicine, Miami, Florida, USA; ⁴Department of Cardiovascular Medicine, Mayo Clinic, Rochester, Minnesota, USA.

Acknowledgments: The authors are grateful for the philanthropic support of our generous anonymous donors, and the philanthropic support of Drs. Mary Ann and Donald A Sens; Mr. Raymond Ames and Ms. Barbara Thorndike; Frank J and Eleanor A. Maslowski Charitable Trust; Joseph F and Mary M Fleischhacker Family Foundation; Mrs. Diane and Dr. Cline Hickok; Mrs. Marilyn and Mr. William Ryerse; Mr. Greg and Mrs. Rhoda Olsen; Mrs. Wilma and Mr. Dale Johnson; Mrs. Charlotte and Mr. Jerry Golinvaux Family Fund; the Roehl Family Foundation; the Joseph Durda Foundation. The generous gifts of these donors to the Minneapolis Heart Institute Foundation’s Science Center for Coronary Artery Disease (CCAD) helped support this research project.

Disclosures: Dr. Sandoval is on the advisory board of Abbott Diagnostics, Phillips, Roche Diagnostics, and Zoll. Dr. Chatzizisis receives speaker honoraria, advisory board fees, and research grants from Boston Scientific; advisory board fees and research grants from Medtronic Inc.; holds US patent (No. 11,026,749) and has international patent pending (application No. PCT/US2020/057304) for the invention entitled “Computational simulation platform for the planning of interventional procedures”; and is the co-founder of ComKardia Inc. Dr. Siontis receives research funding from Anumana, Inc., and is the co-inventor of AI-ECG algorithms, which Mayo Clinic has licensed to Anumana, Inc. with potential for commercialization. Dr. Brilakis receives consulting/speaker honoraria from Abbott Vascular, American Heart Association (associate editor, Circulation), Amgen, Asahi Intecc, Biotronik, Boston Scientific, Cardiovascular Innovations Foundation (Board of Directors), CSI, Elsevier, GE Healthcare, IMDS, Medicure, Medtronic, Siemens, Teleflex, and Terumo; research support from Boston Scientific and GE Healthcare; is the owner of Hippocrates LLC; and is a shareholder in MHI Ventures, Cleerly Health, and Stallion Medical. The remaining authors report no financial relationships or conflicts of interest regarding the content herein.

Address for correspondence: Emmanouil S. Brilakis, MD, PhD, Minneapolis Heart Institute, 920 E 28th Street #300, Minneapolis, MN 55407, USA. Email: esbrilakis@gmail.com; X: @esbrilakis, @m1chaella_alex

Back to August 2024 Table of Contents