Ross Upton, CEO and Co-Academic Founder of Ultromics, updates us on the company’s use of AI in clinical diagnosis and the latest clinical trials using its AI diagnostic tool
Coronary heart disease is the largest cause of death globally – it affects almost 50 per cent of people over 40. Its prevalence is only complicated by its relative elusiveness: simply identifying it can be a major obstacle in any effort to effectively treat it. To save lives, eliminate costs, and make clinicians’ lives easier, it’s critical that it is detected as early as possible.
Currently, when diagnosing the condition, clinicians rely on using an image of the heart as a diagnostic tool, otherwise known as an echocardiogram. They analyse a handful of data points with the naked eye to identify patterns and abnormalities that are characteristic of the disease – eventually arriving at a diagnosis. The problem is this takes time and is entirely reliant on the training and expertise of the operator to reach a diagnosis. When the operator is good then the test is reasonably reliable, but expertise varies significantly – and training clinicians to interpret stress echocardiograms can take a number of years.
If current diagnostic options are unsuitable, technology may provide the necessary solutions to improve conditions for clinicians, hospitals, and patients alike. There is a new method for diagnosing heart disease that uses machine learning, a type of AI, to produce rapid, consistent and expert results. Ultrasound images of the heart contain a lot of information that is difficult to capture using manual interpretation. This information could be used to better identify features indicative of disease and subsequently provide increased diagnostic accuracy.
The AI algorithm ‘EchoGo’ was developed by identifying thousands of data points from each image from one of the world’s largest commercially consented echo database, and then working out how combinations of these data points can be used to predict what happens to the patient over the next year. It’s cloud-based, vendor-neutral software that interprets any echo image taken by any hardware manufacturer – and integrates seamlessly into day-to-day clinical practice.
Because the technology has been developed from information contained within hundreds of thousands of images collected for research over the last 10 years, we have been able to identify patterns from these selected data points that predict coronary heart disease with a high degree of accuracy. Using this method, the algorithm can now reliably and consistently detect the presence of coronary artery disease, which can make misdiagnosis less likely.
Clinical trials are currently underway across both the UK and US to prove this method further. One of these trials, ‘EVAREST’, is expanding to involve 30 NHS sites with a recruitment of 5,000 participants over the next 12 months and ‘Rainer’, based in the US, includes a review of 1,100 stress exams. This is the largest ongoing clinical programme of its kind.
‘EchoGo’ holds promise in improving the entire care pathway: it makes it easier to identify conditions and diseases – meaning more lives can potentially be saved without any disruption to normal clinical workflow. While preventing loss of life and treating patients is any clinician’s priority, it’s worth mentioning that the algorithm will also, if successful, result in potential cost savings. This partially means preventing unnecessary operations, though this is only part of the equation: the other part is preventing patients from being sent home when they have a disease. If the condition isn’t correctly identified and caught, clinicians run the risk of patients having out-of-hospital heart attacks – which can be fatal in some instances, and costly in all instances. They require ambulances and out-of-hours physician consultations which can be expensive.
There may be concern in some quarters that clinicians will eventually be substituted by AI algorithms and other technological advances. This concern is valid, given the prevalence of automation in some other industries and the rapid progress of AI in healthcare – but this is not the case when it comes to our technology.
The purpose of ‘EchoGo’ is to augment clinicians, assist them in making a more accurate diagnosis of disease, and give them greater confidence in these diagnoses – not to replace them. In fact, by augmenting the clinician’s diagnosis, it could reduce training time which may result in the proliferation of more clinicians – as they would have the tool to interpret stress echocardiograms more effectively. Medicine is an essentially human endeavour, and it will only advance if human clinicians are in the vanguard. It’s neither desirable nor advisable for AI to supplant them; algorithms, if nothing else, are not renowned for their bedside manner. With a consistent standard of expertise, and less time spent on training and interpretation, more resources can be put towards helping patients directly.
So, the technology plays a supporting role, but an important one. AI is already being used in healthcare successfully with products like the Butterfly Network’s pocket-sized ultrasound device, which plugs into an iPhone and can image the whole body. The trend of AI-implementation in healthcare is only set to increase, with potential for it to alter the entire paradigm by which the healthcare system operates. AI will not just replace these processes – it will improve them over time. Inefficiencies can be eliminated, costs can be reduced, and lives can be saved with thoughtful, strategic and widespread adoption.
We are continuing to strive for improvement in the diagnoses of all kinds of heart problems with our technology – and are continuing research and development into diagnostic aids for valve disease, heart failure, and other illnesses. If it’s possible to detect these issues earlier, they can be treated earlier: to the benefit of patients, clinicians, and hospital budgets.
For our previous article on Ultromics ground-breaking work, click here.
You're the expert! Write for The Engine or share your articles, papers and researchAdd your content
Add your content
Sign up for Ignition, our regular, ideas-packed newsletter