Routine eye tests may spot blood cancers years earlier, study finds

Routine retinal scans taken during optician visits could help identify people at increased risk of developing blood cancers years before symptoms emerge, according to a new study.

Researchers used artificial intelligence to analyse routine eye scans from more than 1,300 patients in the UK and found microscopic retinal changes that were associated with a higher risk of later blood cancer diagnosis. Compared with people without such changes, those identified by the AI were seven times more likely to be diagnosed with multiple myeloma and twice as likely to be diagnosed with leukaemia within the following ten years.

The study, published in the European Journal of Cancer, examined routine retina images taken by opticians and applied machine-learning methods to detect subtle alterations in the blood vessels of the retina. Senior author Dr. Anant Madabhushi of Emory University said the AI was able to predict the risk of developing multiple myeloma, lymphoma and leukaemia up to ten years before clinical diagnosis.

Blood cancers — which include leukaemia, lymphoma and multiple myeloma — affect about 40,000 people in the UK each year and cause roughly 16,000 deaths annually, making them the country’s third biggest cancer killer after lung and bowel cancer. They are often difficult to diagnose early because common symptoms such as fatigue, night sweats and unexplained bruising can be mistaken for other conditions, and there is no widely used simple screening test.

The authors said chronic inflammation, a hallmark of many blood cancers, may produce microscopic changes to the retinal microvasculature that are detectable with fundus imaging. By training AI algorithms on large numbers of eye scans, the researchers were able to distinguish patterns linked with higher future cancer risk when compared with scans from patients who did not develop blood cancer.

Dr. Richard Francis, deputy director of research at Blood Cancer UK, cautioned that further studies are needed before the approach could be implemented in routine clinical practice. "While more research is needed before this could be used in clinical practice, these findings provide an important proof of principle that AI-driven tools may one day help us intervene earlier and improve outcomes," he said.

The investigators emphasised that their results do not establish a diagnostic test for blood cancer but suggest a potential screening pathway using already-widely available retinal imaging. If validated in larger, prospective cohorts and diverse populations, an AI-based flagging system in community optician clinics could prompt earlier specialist referral and blood testing for people at elevated risk.

Limitations of the current analysis include its retrospective design and the need to confirm findings across different imaging devices, populations and health-care settings. The study also compared scans from patients who later developed blood cancer with those who did not, but it did not determine whether intervening earlier based on retinal findings would change survival or other clinical outcomes.

Researchers and clinicians said the next steps should include prospective trials to assess AI performance in routine screening environments, evaluation of cost-effectiveness, and strategies to integrate any effective tool into existing care pathways without causing undue anxiety or unnecessary testing for patients.

For now, the study adds to growing evidence that noninvasive imaging combined with machine learning can reveal systemic disease signals, and it highlights the potential of routine eye exams as a low-cost platform for broader health screening if subsequent research confirms the early results.