'AI helped diagnose my daughter's rare condition'

The family of a 19-month-old girl with a rare genetic condition affecting only six people in the UK say AI helped them to gain a diagnosis.

Lily, from Bath, had surgery for an unknown heart condition at four months old. In January tests finally revealed she has Multisystemic Smooth Muscle Dysfunction Syndrome (MSMDS).

Her mum, Rosie, said she put Lily's symptoms into ChatGPT, which came up with MSMDS, but when she told doctors, "some of them had never heard of it; others said it is so rare it's very unlikely".

Nick Meade, chief executive at Genetic Alliance UK, said he would "urge caution" using ChatGPT as rare diseases are particularly affected by AI modelling problems.

Research published by University of Oxford earlier this year echoed Meade's warning.

Dr Rebecca Payne, who co-authored the study, previously told the BBC that research found "despite all the hype, AI just isn't ready to take on the role of the physician".

"Patients need to be aware that asking a large language model about their symptoms can be dangerous, giving wrong diagnoses and failing to recognise when urgent help is needed," Dr Payne, who is also a GP, added.

MSMDS is caused by a specific change in the ACTA2 gene and affects organs such as the heart and kidneys.

As well as the heart condition, Rosie said Lily's pupils were always dilated. She now needs to wear sunglasses when outdoors.

Lily is one of only 70 people known to have the condition worldwide.

She is under the supervision of Great Ormond Street Hospital and Bristol Royal Hospital for Children.

The BBC has contacted the hospitals, but neither has responded.

Lily's parents, Rosie and Jonny, have set up a charity, ACTA2 Alliance UK, to fund research into the condition.

They said international research into potential treatments and a cure is being led from Boston, America, while other promising approaches are also being explored in the UK.

"We are speaking to some people in England about research into other treatment options, so that's one of our current targets of £30,000 to raise towards that for a proof of concept, so we're really hoping that can be another option, alongside the incredible work already happening internationally," Rosie added.

This week, the family held a move-a-thon to help raise money for research and is asking people to log every mile they step, dance, swim, walk, or run to help cover the equivalent of the 3,300-mile distance from Bath to Boston.

Lily's grandfather, Andy, said they also wanted to raise awareness of the condition to "identify patients to build up data".

"We feel we are on a bit of a ticking time bomb because it is a progressive condition. We're on a journey to find a cure, and we're funding vital medical research that could lead to that," he said.

Meade stressed that Lily's case was rare and Genetic Alliance UK is not aware of many diagnoses being made as a result of AI.

"While we welcome AI's potential to improve outcomes for families, we'd also urge caution. Machine learning models learn mainly from whatever appears most often in their training data," he said.

"This is a well-documented problem in AI, and rare conditions are, by definition, the kind of case it affects most."

NHS England was approached for comment.

Follow BBC Somerset on Facebook and X. Send your story ideas to us on email or via WhatsApp on 0800 313 4630.