5 April 2024
An exciting new study aims to use artificial intelligence and machine learning to identify patients with scleroderma who are the most likely to develop severe lung disease, allowing for earlier treatment.
Professor Elizabeth Renzoni, respiratory consultant at Royal Brompton and Harefield hospitals, and her team were the sole recipients of a joint funding scheme from Scleroderma and Raynaud’s UK (SRUK) and the World Scleroderma Foundation, which awarded a single project under a ‘winner-takes-all’ scheme.
The funding will support a project which aims to use AI to predict outcome in a type of lung disease known as systemic sclerosis-associated interstitial lung disease (SSc-ILD).
Scleroderma and interstitial lung disease
Scleroderma, also known as systemic sclerosis (SSc), is a rare autoimmune disease where an individual’s own immune system attacks the connective tissue in the body.
The disease causes fibrosis (hard thickened scar-like tissue) of the skin and sometimes of internal organs including the lungs, digestive system, heart or kidneys.
The type of lung disease patients with scleroderma develop is known as interstitial lung disease (ILD). ILD occurs in a large proportion of patients with scleroderma and is often abbreviated as SSc-ILD.
The interstitium of the lung is the tissue surrounding the air sacs and blood vessels. In ILD this interstitium is fibrotic (scarred) which makes it thick and stiff. This makes it harder to breathe.
Scleroderma patients can also develop pulmonary hypertension, where the walls of the vessels in the lung become thicker, with increased pressure of the pulmonary artery and eventually a strain on the right side of the heart.
ILD and pulmonary hypertension are the main cause of death in patients with scleroderma.
Progressive disease
How SSc-ILD develops over time varies considerably between individuals.
In some patients the lung fibrosis remains mild but approximately a third of patients will go on to develop progressive ILD, where there is an increasing amount of lung scarring, worsening breathlessness and reduced survival. Some patients with SSc-ILD will also develop pulmonary hypertension, which further shortens survival.
There are a few medications which can prevent or slow the worsening of lung fibrosis, but these come with side effects.
If the disease progression could be predicted in these patients, then those with mild lung disease would not need to take the medication (or receive a lower dose), whereas those likely to develop severe lung disease in the future could be put on medication sooner.
Likewise, it would be useful to predict early on which patients will develop pulmonary hypertension, as earlier treatment could lead to better survival.
Machine learning research
Machine learning is a type of artificial intelligence (AI) where a computer develops a set of rules called algorithms, which it learns from a set of information.
In this project, machine learning will be used to develop computer algorithms to assess disease features shown on high resolution CT scans of patients with SSc-ILD, alongside data from lung function tests.
The reason for using machine learning algorithms is because it is difficult to quantify disease progression in imaging scans by eye, even for trained radiology specialists. This includes patterns not easily detected by the naked eye, which predict which patients are going to develop progressive SSc-ILD or pulmonary hypertension.
The overall aim of the study is to be able to identify which SSc-ILD patients are most likely to go on to develop severe lung disease.
This will allow for earlier treatment to slow down the lung fibrosis and help them to live longer with milder symptoms.
Dr Carmel Stock, a research associate working on the project, explained the importance of the research for this group of patients, and said:
“Patients with SSc are in need of tools that can reliably predict the development of progressive lung disease. If successful, this study has the potential to reduce anxiety and uncertainty in patients on the progression of their condition, as well as the ability to introduce treatments earlier to prevent or slow down progression.
“Since implementation of the machine-learning algorithms in this research does not require specialist equipment, it can easily be integrated into existing hospital systems. This would provide low-cost decision support to referral centres, and in institutions where expertise in this condition is limited.
“Ultimately, this project has the potential to have an immediate impact in care for patients with SSc-ILD.”
To find out more about our research, please contact us.
Read more research stories or sign up to the research newsletter.