Currently working up to proof-of-concept stage, the technology being developed by Demetria aims to speed up diagnosis, improve patient outcomes, and eliminate the need for laparoscopic surgery currently used to provide a definitive diagnosis of the disease.
Endometriosis is a chronic condition impacting over 1.5 million women in the UK, and nearly 200 million globally. In the UK alone, endometriosis costs the UK economy around £8.2bn every year in healthcare outlay. It occurs when tissue similar to the lining of the uterus grows outside of it, which can lead to severe pain, infertility, and other complications.
Diagnosing the condition is challenging because its symptoms overlap with other health issues such as fibroids.
“The average woman in the UK takes more than 10 ultrasound visits and 10 gynae appointments to even get to a diagnosis,” explained Lorna Maclean, Demetria’s founder and CEO. “And then, currently, the only definitive gold standard for a diagnosis is an invasive laparoscopic surgery, which a lot of women don't want because there's no cure for this.”
Maclean brings her lived experience to the development of the AI-enabled diagnostic tool, both as someone living with endometriosis and as someone who has worked in an AI-related field and realising that diagnosing the condition was ‘a really beautiful machine learning problem to solve’.
Despite frequent visits to her GP, it took years before Maclean was offered any form of diagnostic imaging. UK guidelines recommended a transvaginal ultrasound scan, but this option is only available to women aged 16 and over.
Instead, she was given abdominal ultrasound scans, which have low accuracy in detecting endometriosis, especially if the disease is not yet widespread.
“If it does then you're already in trouble, because it's extended up into the abdominal rather than just gynaecological anatomy,” she said.
MORE FROM MEDICAL & HEALTHCARE
Even with access to a transvaginal ultrasound later in life, Maclean discovered that this method was only about 47 per cent accurate unless performed by an expert with years of specialised training and there are very few radiologists worldwide with such expertise.
“It's a very difficult machine learning problem to solve because the disease can present in so many different ways, and often it's minor cell changes,” said Maclean.
The AI-enabled solution is designed to be integrated with existing ultrasound hardware, making it accessible and cost-effective for widespread use.
In the first phase of development, Demetria’s solution will assist radiologists by analysing still images from ultrasounds uploaded into the company’s software. The system will act as a ‘co-pilot,’ guiding radiologists to potential areas of concern by flagging minor cellular changes indicative of endometriosis.
Maclean’s team then plans to advance the technology further, incorporating video scanning, which can capture movement and adhesions between tissues, providing even greater accuracy.
“Sometimes things are just so stuck together, almost like chewing gum, when you're doing an internal ultrasound scan that's indicative of endometriosis,” said Maclean. “You won't necessarily pick that up on a still image, but you will have a much better understanding on a video.”
Demetria is being helped on its journey by impulse, Cambridge University’s entrepreneurship programme.
“impulse has given me the tools and connections to turn an idea born of frustration into a viable solution,” said Maclean. “It is quite remarkable that such a programme exists, where so many of the UK’s most successful entrepreneurs are willing to offer their time so generously to young startups.”
AI is a gamble we cannot afford without cybersecurity
I am reminded of a quote and example from palaeontology. Herbivorous dinosaurs like stegosaurs/triceratops originally weere bi-pedal with large...