A new artificial intelligence (AI) tool could soon help general practitioners (GPs) detect lung cancer in patients up to four months earlier than current methods. This could significantly improve chances of survival and lower the costs of treatment.
Developed by researchers at Amsterdam UMC in the Netherlands, the algorithm was created using data from over 500,000 patients and is designed to work during regular GP visits. The findings of this study were recently published in the British Journal of General Practice.
Lung cancer is one of the most deadly cancers worldwide. In many cases, it is only diagnosed at a late stage, when treatment is less effective. Around 80% of patients diagnosed in the later stages die within a year.
Earlier detection has already been shown to improve outcomes. For example, getting diagnosed just four weeks earlier can make a noticeable difference. A four-month lead could therefore have an even greater impact.
The new algorithm looks through the entire medical history of a patient—including the GP’s handwritten notes, which are usually unstructured and difficult for computers to analyze. While many existing tools only use fixed data points like “smoker” or “coughing up blood,” this AI system picks up subtle clues from the notes that a GP has written over the years.
These could include repeated mentions of mild symptoms, patterns in appointments, or other small details that wouldn’t trigger immediate concern on their own.
Professor Martijn Schut, one of the lead researchers and a professor of Translational Artificial Intelligence at Amsterdam UMC, explained that the algorithm can spot signs of cancer based on this long-term data.
In fact, it was able to identify a large number of patients who were later diagnosed with lung cancer—up to four months before they would normally have been referred for testing.
The study used information from four academic general practice networks in the Netherlands—Amsterdam, Utrecht, and Groningen—covering the medical records of more than half a million patients. Out of these, 2,386 were diagnosed with lung cancer.
These diagnoses were confirmed using the Dutch Cancer Registry. The algorithm was trained using both structured data (like medical codes) and unstructured data (like free-text notes from GPs).
Compared to national cancer screening programs, this AI tool may have some key advantages. One is that it can be used directly in a consultation, without the need for a separate test or scan. It also appears to produce fewer false positives—cases where the system wrongly suspects cancer—than some traditional screening methods.
Beyond lung cancer, the researchers believe this method could be adapted to detect other hard-to-spot cancers, like pancreatic, stomach, or ovarian cancer. These types of cancer also tend to be diagnosed at late stages, making early detection especially important for improving survival and quality of life.
While the findings are promising, the researchers stress that more work needs to be done before the tool can be used in everyday practice.
For example, more studies are needed to understand exactly which parts of a patient’s medical history the algorithm is using to make its predictions. The system also needs to be tested in other countries and health care systems to ensure it works just as well outside the Netherlands.
Overall, this study shows how artificial intelligence can be used to support doctors and improve patient outcomes.
By using all available patient information—including past visits and handwritten notes—the tool gives GPs a better chance to catch cancer early, when it is more treatable. If future tests confirm its accuracy, it could become an important part of everyday health care.
If you care about lung health, please read studies about marijuana’s effects on lung health, and why some non-smokers get lung disease and some heavy smokers do not.
For more information about health, please see recent studies that olive oil may help you live longer, and vitamin D could help lower the risk of autoimmune diseases.
The research findings can be found in British Journal of General Practice.
Copyright © 2025 Knowridge Science Report. All rights reserved.
