Exploring the role of artificial intelligence (AI) in heart disease

From unlocking our phones with Face ID, to using smart home devices to turn lights on, or asking ChatGPT to craft the perfect work email, artificial intelligence (AI) is infiltrating every aspect of our lives.

With advancements in these technologies skyrocketing, what could an AI-focused future mean for how we treat and manage heart disease?

This is a subject of great interest to Dr Nikki Bart, cardiologist and Head of the Institute’s Infiltrative Cardiomyopathy Lab, who is being recognised globally as a leader in the field of AI and journal authorship. We caught up with Dr Bart to get her thoughts on how AI is likely to impact cardiology, patient care, and the role of the clinician.

What are some of the AI advancements we are seeing, or are likely to see, in cardiology in the coming years?

AI is getting involved in many aspects of medicine - not just cardiology. This is starting right from when you apply to study medicine. For example, you could say, ‘Hey ChatGPT, can you write me an application for medicine? Here are my credentials’. One of the issues we are dealing with now is how can we determine what's a student-written application or assignment versus what's written by generative AI like ChatGPT?

Then in the junior doctor years, the basic note taking can be streamlined with AI – which I think is fantastic because the more we can streamline processes that are prone to human error, the better. There are programs now for medical note-writing that's AI-driven - they're not quite at the standard we need them to be yet, but with a bit more work they will likely get there.

Then we move into areas like diagnostics where AI models can excel. For example, in cardiology, an AI model has been shown to diagnose a type of heart attack known as ST-elevation myocardial infarction (STEMI) as well as diagnosing certain types of arrhythmias. There are also a number of echocardiogram (echo) parameters that can be read better by AI than sonographers and cardiologists. We are seeing similar results with MRIs, where AI’s ability to quantify volumes and ejection fraction and other such parameters is quite amazing. So AI can be beneficial across the whole spectrum of medicine.

What are some of the positives and negatives of AI advancement in cardiology?

I think there's a lot of potential positives for cardiology, and medicine at large, that come with AI – and I am really interested in this space because I want to improve patient care and patient outcomes. I think you've got to get on the bandwagon with AI, because it's definitely here to stay, and we've got to do it in such a way that we always have the patient at the centre.

In terms of improving patient outcomes with AI, there’s no question that there's many aspects of cardiology could be automated to improve efficiency and reduce the likelihood of human error. For example, if I was to enter values into an Excel spreadsheet, my likelihood of committing human error is quite high, whereas if you design a model to do that, then we can avoid those errors.

In terms of the negatives in the field of medicine, I think we've got to be careful about data regulation and data privacy. Most health records are now electronic, and there's lots of good research that for rare diseases, if you combine an electronic health record with ECGs, genetics and other imaging, then you can identify rare diseases earlier. This is great for patient outcomes, but then it brings up the question, ‘how do you keep that medical record safe and keep patients safe in terms of their data privacy and their data security?’. The other important area is that of accountability. If a doctor makes a mistake, we can learn from it and correct the human and system errors. If an AI algorithm makes a mistake, who is accountable? How do we learn from the mistake if the algorithm is a black box? Other concerns are data integrity and data ownership.

What role do you see cardiologists and other clinicians playing in an AI-focused future?

In an AI-focused future, I see the role of cardiologists and other clinicians as being focused on ensuring the patient always comes first. The example that I used in an article I wrote last year is of a patient with advanced heart failure. Can AI diagnose that patient? Maybe. But can it sit with that patient? Can it ask that patient ‘what are you worried about and what do you care most about?’ – which is important to patient decision-making. AI doesn’t have the nuance to understand that for a patient their main concern for that day might be getting home to a family for a meal or getting out in the sunshine for a walk. I don't know how you could build that into an algorithm.

I really hope that with regulation and care for how these AI models are applied that yes, they will improve our efficiency, they'll cut back on the time we spend on paperwork and all that checking and double-checking. Hopefully, AI can make these processes more streamlined so that we've got more time to sit down with the patient and say, ‘what's the most important thing to you today and how can I help you today?’. That way we can really walk with that patient from diagnosis to treatment to care.

What are some of the ethical considerations we need to keep in mind as we incorporate AI technologies into the management of heart disease?

One area of ethical consideration that I'm really fascinated in is AI and journal authorship. AI is getting good at writing medical literature, including journal articles, and the more we give it in terms of input, the better it is at writing these niche areas. So should AI be classed as an author? Our current criteria for authors comes from the International Committee of Medical Journal Editors (ICMJE) and requires substantial contribution, drafting the work critically, final approval of a manuscript and accountability. The large language models (LLM) such as ChatGPT do not meet these criteria. So I think it's really important that we hold the medical literature to the highest standard, which is that AI does not yet qualify to be an author.

The reason why this is important is that for clinicians, and clinician-scientists like me, we need to be able to see a patient and what they’re presenting with and then ask, ‘what does the literature say here?’ so we can ensure we’re practising evidence-based medicine. For example, if the most recent trial says one treatment is the most effective treatment, you can then discuss that option with the patient. If there's no accountability in that literature, that brings up some serious ethical and moral questions about the information we are using in our day-to-day practice.

As a clinician and researcher, what interests you most about AI and its potential to manage heart health?

Cardiology is fast-paced, and one of the fantastic things about being a cardiologist is that there are regularly many technological advancements- and AI is no exception. There’s this interaction between the fast pace of technology, the ability to be more efficient, and patient care - and I think what interests me the most is being able to embrace the power of it, while at the same time making sure there's all the regulations and checks and balances to make sure that we're safe with it.

So I think as a cardiologist it’s that real interface between, here's what the science is saying and here's how we can help patients – it’s a really beautiful interface that’s really key to ensuring we can treat and protect the heart health of our patients.

What would you say to people who are wondering how to navigate AI advancements as a patient?

I can understand that for many patients the thought of navigating this new unknown AI future could be daunting. That being said, I really hope that the clinician-patient interactions mean that we can provide information and reassurance to people that we are still having a patient-centric approach and that we are holding ourselves to the highest standards.

Technological changes like AI are often fast paced, which can be concerning for patients, and so I would want to go through those concerns with my individual patients and answer questions wherever possible. But I think as with any new field, it's reasonable to have a degree of anxiety and we've got to manage that and do our best to allay some of those fears where possible so patients feel more confident as they navigate the AI advancements to come.