AI urine analysis predicts infection in lung disease patients a week in advance

Analysing urine samples with artificial intelligence (AI) can predict when patients with a chronic lung disease are likely to have a flare-up seven days before symptoms start, a study has found.

The technology could help personalise treatment and prevent hospitalisations, academics said.

The research involved patients carrying out a simple daily dipstick test on their urine – similar to a lateral flow test – and sharing the results with experts using their mobile phone.

For the study, researchers analysed urine samples from 55 patients with chronic obstructive pulmonary disease (COPD) to pinpoint how molecules change when symptoms worsen.

COPD is a term used for a group of lung conditions that cause breathing difficulties such as emphysema and chronic bronchitis.

The current treatments are reactive to a severe illness. It would be better if we could predict an attack before it happens and then personalise treatment to either prevent the attack or reduce its impact

Symptoms can include shortness of breath, wheezing and a persistent chesty cough.

Flare-ups, also known as exacerbations, happen when symptoms suddenly get worse, and are most common during the winter.

Professor Chris Brightling, of the University of Leicester, who led the study, said: “COPD exacerbations are when someone with COPD becomes very ill and needs additional treatment either at home or in hospital.

“The current treatments are reactive to a severe illness. It would be better if we could predict an attack before it happens and then personalise treatment to either prevent the attack or reduce its impact.

“We wanted to develop a predictive test that would act like a personal weather forecast of an impending flare-up.”

After identifying the changing molecules, researchers developed a test to measure the levels of five different biomarkers in urine.

Some 105 patients with COPD then tested their urine every day for six months with the dipstick test and shared their results with researchers.

The advantage of sampling urine is that it’s relatively quick and easy for patients to do at home on a daily basis

The results from 85 were analysed using an artificial neural network (ANN), which is a type of algorithm that uses a network of artificial neurons to process data in a way that mimics the human brain.

The study, published in ERJ Open Research, found the AI model could accurately predict a flare-up up to seven days before symptoms started.

Researchers acknowledged the study had a number of limitations, including a small sample size.

Prof Brightling added: “The advantage of sampling urine is that it’s relatively quick and easy for patients to do at home on a daily basis.

“We need to do more work to refine the AI algorithm with data from a bigger group of patients.

“We hope this will allow us to create AI testing for COPD patients that will learn what is ‘normal’ for each person and forecast a flare-up in symptoms.

“Patients’ care could then be adapted, for example they might need further testing or treatment, or they might be able to limit their exposure to triggers like pollution or pollen.”

Reacting to the study, Dr Erika Kennington, head of research and innovation at the charity Asthma + Lung UK, said: “This quick and non-invasive test shows how our urine could be used as a warning of worsening lung health.

“Allowing people with chronic obstructive pulmonary disease to take steps to manage their condition before it gets worse could really help people stay well and out of hospital.

“However, this compelling research would next need to be tested in a much larger group of people with COPD and the cost-effectiveness analysed, before it could be used in a health care setting.”