Skip to main content

From potential to practice: using AI to improve cardiovascular diagnosis

(Image credit: Image source: Shutterstock/everything possible)

Artificial intelligence (AI) is often discussed in the future tense, as though it’s a far-off technology that hasn’t yet arrived. The truth, however, is that it’s not a future concept, as it’s already arrived in many arenas, including healthcare. A recent report from PwC identified several current implementations of the technology. With the assistance of AI, clinicians can more easily and accurately review and process mammograms, eliminating unnecessary biopsies from false positive diagnoses. AI-based technology is already helping surgeons complete complex operations and are assisting clinicians on a day-to-day basis. Big data analysis is allowing physicians to identify at-risk patients – helping to ensure they receive treatment before minor problems mutate into major life-threatening conditions – and provide crucial, early support for patients who want to stay healthy.

Specific AI implementations currently in use include products such as Butterfly Network’s pocket-sized Ultrasound, which can image the entire body, and Google’s DeepMind Health, currently in partnership with Moorfield’s Eye Hospital in London, to develop machine-learning based AI technology to address the issue of macular degeneration in aging eyes.

In some form or another, AI is being used all across the medical profession: from detection to treatment, to palliative and end-of-life care. In cardiovascular health, current and future implementations of this technology may have particularly exciting implications for patients, clinicians, and the entire medical profession.

The challenge with a cardiovascular diagnosis

Coronary heart disease is one of the largest causes of death globally – some 50 per cent of people over 40 are affected, and it is the single most common cause of death in the UK. 2.3 million people are currently suffering with it, and those afflicted are twice as likely to have a stroke. In order to effectively treat this and other cardiovascular diseases, swift detection is key.

But simply identifying these diseases is often complicated. When a clinician diagnoses a cardiovascular disease, they rely on an echocardiogram – a scan of the heart which is manually analysed by a cardiologist or clinician. It’s the most widely used tool for diagnosing heart disease with over 60 million undertaken annually in the US alone. An echocardiogram utilises sound waves that create a scanned image of the heart. When looking at this image, cardiologists may determine five to ten factors or markers and base their diagnosis on these.

The clinician identifies patterns and notes any abnormalities that signal disease, eventually arriving at a diagnosis. However, this can take time, and as with any diagnosis, is reliant on the training and expertise of the clinician. Overall, the current process of diagnosing heart disease yields a detection success rate of 80 per cent – but one in five cases goes misdiagnosed. Even the best-trained clinicians are limited by what they can see with the naked eye, so patients sometimes have to undergo unnecessary surgical procedures or be sent home with the disease, both of which can be fatal.

How AI can revolutionise cardiovascular health

Computational technologies and diagnostic algorithms may hold the key to the future of cardiovascular diagnosis. There is real potential to increase the diagnostic accuracy of echocardiograms by using AI. At Ultromics, we’ve developed some of the most accurate echocardiography software in the world to prove it.

Our machine learning software has already increased the accuracy of echocardiogram testing from 80 per cent to 90 per cent. By incorporating AI with machine learning technology into these tests, thousands of markers and data associated with various cardiovascular diseases can be cross-analysed and assembled into patterns. This makes diagnosis more accurate for experienced clinicians, is more efficient, and can save more lives in the process.

Support, not substitution

There may be some concern that AI technology and other technological advances will eventually substitute human clinicians and cardiologists. It’s an understandable concern, especially with the rise in automation and the rapid progress in AI technology.

But the worry is somewhat misplaced. In the field of cardiology, machine learning AI software algorithms will not replace clinicians, but augment them. This may improve the quality of diagnosis and reduce training time as well. Educating clinicians to accurately and efficiently interpret echocardiogram images can take years under normal circumstances; with AI, this training period can be considerably shortened.

The upshot of all this is that, with the aid of AI technology, clinicians will have greater confidence in their diagnoses. It’s about eliminating ambiguity and integrating new algorithms and software into existing echocardiogram systems and medical workflows. It supports the clinician, rather than substituting them, and ensures that diagnosis and analysis are undertaken with a consistent, effective methodology.

Cold, heart facts

Artificial intelligence (as opposed to human intelligence) may receive mass media attention but the reality is that the majority of applications are just smart pieces of software capable of providing clever statistical analysis. As impressive or daunting the media claims AI to be, the most tangible benefits of AI may come from utilising the software for detection and prevention, improving systems and reducing operational burdens, which will ultimately empower clinicians to do their best work. In our lifetime I do not expect to see robot doctors, walking around hospitals prescribing medications.

It may also make physical health technology products smaller – to the point where the coronary heart disease patients of the future could receive their diagnosis on their smartphones. This relieves a lot of the pressure on clinicians, and ensures that when they make an appointment, it’s with a view to developing a treatment plan rather than establishing the medical problem. If it comes to pass, it may fundamentally change first-line cardiology diagnosis across the entire healthcare profession. The patients of the future could very well find out if they have heart disease in their homes, rather than their hospitals.

But that’s some way away. In the meantime, small increases in diagnostic accuracy can potentially save many lives, and a 10 per cent increase is far from small. Smart AI-enabled software can help clinicians catch coronary heart disease, but it can also take a weight off their shoulders – allowing them to focus on advice, support, and expert treatment. Best of all, it’s just getting started: with time, investment, and widespread adoption, the technology will only get more sophisticated. AI has potential applications in many particular areas; right now, though, it seems to be highly effective when it stays close to the heart.

Ross Upton, CEO and Co-Academic Founder, Ultromics
Image source: Shutterstock/everything possible

Ross Upton is the CEO and Co-Academic Founder of Ultromics.