Home » Transforming healthcare » Imaging » AI can be an evolution, not a revolution
Imaging

AI can be an evolution, not a revolution

AI robot surgical
AI robot surgical
avatar

Hugh Bettesworth

Chief Executive at Mirada Medical

The use of AI in imaging need not be confined to radical changes in healthcare delivery. It can also be used with existing systems to save time and improve patient outcomes.


Much has been made of how AI could change the face of healthcare – and many have expressed concern about such a disruptive revolution in the clinical world. However, applied to existing, everyday healthcare tasks, AI can make a difference – without the wait for clinical trials required to validate major technological innovations.

So says Hugh Bettesworth, Chief Executive at Mirada Medical, a company that specialises in software for radiation oncology and diagnostic imaging.

Radical new technology has to be validated by time-consuming clinical trials

“There are many innovative, deep learning (AI) technologies being developed to aid diagnostic imaging, but radical new technology has to be validated by clinical trials to prove that it is at least as effective as an experienced reading radiologist. These processes take time, which means the patients, clinicians and the NHS must wait many years for the benefits to be delivered,” says Bettesworth.

However, he says, AI can – and increasingly will – be used to improve the efficiency of existing techniques without the need for revolutionary technology.

“AI can also be applied to the existing technology used in many simple, everyday healthcare tasks. There is a whole range of day-to-day healthcare problems where AI technology can be applied today,” he says.

For instance, he points to the value of using AI in organs-at risk-contouring – the process of delineating on an image the physiological structures and healthy tissue that need to be avoided when a radiotherapy beam is targeting a tumour.

Improving the accuracy of patients’ treatment plans

“Good contouring improves patient outcomes because it improves the accuracy of the treatment plan and its ability to avoid, as far as possible, the irradiation of areas of healthy tissue around tumours,” says Bettesworth. “However, contouring is very time consuming and studies show significant inconsistency in the accuracy of contours drawn.”

A pre-treatment CT scan is carried out and the radiation oncologist uses drawing software to draw round the healthy structures to be avoided. “This can take up to two hours and it is tedious work,” says Bettesworth. “Alternatively, contouring is carried out automatically by a software programme, but as these use older technology, they do not always produce contours that compete with those drawn by the most experienced radiation oncologists, and thus require time-consuming editing.”

Dr Mark Gooding, the Chief Scientist who was closely involved in the development of Mirada Medical’s DLCExpert™ technology, which uses AI to learn the user’s contouring preferences and automatically apply them to images, says: “This improves consistency and saves time for the radiation oncologist, while still giving them the chance to amend the contouring themselves. Improved contouring could lead to a reduced chance of side-effects or recurrence for patients.”

He adds: “This shows that AI is not just about huge technological leaps forward. It can be more rapidly applied to everyday tasks to make incremental improvements to the effectiveness of treatment planning, saving time for oncologists and potentially improving patient care.”

Next article