With approx 18 million victims per yearcardiovascular diseases are leading causes of premature death on a global scale, far ahead of cancer and diseases of the respiratory system. Most of these deaths are the result of a heart attack or stroke.
Finding ways to prevent them is therefore a top public health priority. And that’s what British researchers are trying to do using a new tool based on machine learning.
AI, a powerful ally for diagnosis
For several years now, the potential of artificial intelligence has become increasingly apparent. In addition to the obvious potential in research, for example in the development of new drugs, it is also beginning to establish itself in the clinical field. This especially applies to diagnostic science. For example, doctors today can rely on algorithms that can analyze X-rays or MRI images to detect anomalies, such as tumors (see our article).
This is a promising technology, as these images are often very difficult for people to interpret. Even the greatest experts can miss certain elements. These AI-based systems can therefore improve the accuracy and speed of diagnosisand thus the rest of the support.
The latest work by a team of researchers from the prestigious University of Oxford also uses this concept, with one difference. Instead of looking for tumors, they want to watch for warning signs of a heart attack.
Algorithm with hand on heart
In order to determine the risk of a heart attack, doctors often use the so-called test CT scan (scanner in everyday language). This imaging technique allows the recognition of various changes that indicate an increased risk. We can quote narrowing of the coronary arterieswhich is often a very telling warning signal.
The problem is that more than three-quarters of patients who experience a heart attack have no obvious narrowing. The whole challenge is therefore to succeed use other parameters to improve diagnostic accuracy. But it’s much easier said than done. Indeed, these health concerns often stem from a multitude of factors that are difficult to identify. To achieve this, Oxford researchers relied on AI.
They developed a tool based on machine learning which analyzes CT images to detect subtle changes. During this work, they were mainly interested in the changes in the fatty tissue surrounding the arteries suffering from inflammation.
This algorithm was trained on thousands of images taken from patients who later suffered a heart attack. The program learned from these sources precisely identify these risk factors. The first tests proved conclusive; after the first phase of training, the algorithm showed that it was possibleanticipate a heart attack ten years before it happens.
Clinical trial full of promise
The researchers therefore decided to conduct a clinical trial testing the system on 744 people. For each algorithm, they sent several scanner images. These results were then sent to the patients’ doctors for informational purposes, so they could take them into account in their care.
And the result was quite spectacular. In light of these new data, doctors decided to change treatment in 45% of cases ! This shows that the information provided by the algorithm was valuable to caregivers and thus to their patients.
And the icing on the cake is that the researchers think the system can still be improved. It would be enough to keep feeding it new images to improve its accuracy. Ultimately, it could be he refused to monitor other serious cardiovascular diseases. This includes strokes.
A tool that could already make a difference
But they also think he could they already represent a great added value in terms of public health. For that, it would have to be used to a large extent inside National Health Servicethe center of the British health system.
For the authors, this could significantly improve the diagnosis of patients who start experiencing chest pain. This tool could also work wonders in terms of prevention, recognizing risks before these symptoms appear. If necessary, this would allow those affected to adjust their lifestyles to reduce their risk.
They believe this strategy could reduce the number of heart attacks by 20%, and reduce the number of people who die from it by 8%. That would be a huge success in terms of public health.
“ Too many people die from these heart attacks every year », laments Sir Nilesh Samani, director of the British Heart Foundation. ” Harnessing the potential of artificial intelligence to guide patient care is vital. It is hoped that this technology will be introduced into the NHS, preventing thousands of avoidable deaths every year. »