Could AI help doctors identify lung cancer earlier? New UK study shows promise
A new UK study has found that artificial intelligence (AI) could help doctors identify which lung nodules are more likely to be cancer.
Researchers analysed 2,071 CT scans from 1,026 people across seven UK hospitals, using machine learning to examine features of lung nodules that may not be obvious when looking at a scan.
The findings, published in European Radiology Experimental on 5 October 2026, suggest the technology could help doctors make more informed decisions about which nodules need further investigation.
What are lung nodules?
Lung nodules are small areas of abnormal tissue that can appear on a chest scan.
Most lung nodules are not cancer, but some can be an early sign of lung cancer.
Doctors currently assess factors such as the size and appearance of a nodule, alongside other information, to decide whether further scans or tests are needed.
What did the LIBRA study find?
The research, known as the LIBRA study, involved researchers from Imperial College London, the Royal Marsden and other UK hospitals.
They developed a machine learning model that examined detailed features of lung nodules on CT scans.
The best-performing model was better at distinguishing cancerous from non-cancerous nodules than the existing risk assessment methods used for comparison.
Researchers found that the model could potentially help doctors assess lung nodules more accurately, although further testing is needed.
Could this help people who have never smoked?
Lung cancer can develop in anyone, regardless of smoking history.
People who have never smoked are not routinely eligible for England's lung cancer screening programme, which focuses on people with a smoking history who meet specific age and risk criteria.
Although this study was not specifically focused on people who have never smoked, it looked at nodules found outside screening programmes.
Improving how these nodules are assessed could help doctors recognise lung cancer earlier, regardless of someone's smoking history.
What happens next?
The findings are promising, but the AI model is not yet ready for routine NHS use.
Researchers say further testing is needed to confirm how well it works in different clinical settings and whether it improves decisions about lung nodule investigations.
For now, the study offers a possible way to improve lung cancer diagnosis in the future.
Read the research
Hunter B, et al. European Radiology Experimental. Published 5 October 2026.
