In recent years, a new tool powered by artificial intelligence (AI) has started to transform how lung cancer is detected in Croatia. The program called Sybil is now part of the national screening effort, and it is already making a difference.
What is Sybil?
The name Sybil comes from a Greek word meaning “prophetess” or “seer.” In medical context, that is quite fitting: the AI “foresees” which scans carry a higher risk of lung cancer developing.
Technically, Sybil is a deep-learning model: it analyses a particular kind of scan called a low-dose chest CT (LDCT), where patients are exposed to minimal radiation, and predicts their risk of developing lung cancer over the next several years.
What makes Sybil special: even from a single scan, it can estimate the chance a patient might get lung cancer in 1 to 6 years. That is unlike traditional screening which only looks for obvious nodules or changes at the time of scan.
Real practice and how it is used in Croatia
Since October 2020, Croatia runs a national lung-cancer screening programme using LDCT + AI. People eligible (typically older adults with a history of heavy smoking) are referred by their family doctors through a digital platform, no paper referrals needed.
There are 27 screening centres across the country, so you do not have to go to Zagreb, you can do the CT scan in a local centre and the image is sent digitally to specialists. On a daily basis, medical staff review around 100 CT scans from all over Croatia with the help of Sybil.
After the AI “flags” scans with higher risk, specialists (radiologists, pulmonologists) examine them thoroughly to decide if further tests or treatment are needed. The AI does not make the final call, it helps doctors prioritize and spot what might be missed if many scans had to be visually checked under time pressure.
Early results and impact so far
From October 2020 through October 2025, the programme has conducted over 74,000 low-dose CT (LDCT) scans among 52,000 participants.
As of 2025, the screening has identified nearly 700 cases of lung cancer. Of those detected via the programme, more than 50% of patients have undergone surgery, meaning cancers were discovered at a stage early enough for operative treatment. According to a 2025 report, in those detected cases the share of cancers found in early stage (sufficiently small and operable) exceeded 30%.
What to keep in mind!!
Sybil does not give a 100% guarantee. It calculates risk, but every suspicious case still needs human follow-up. Some flagged changes might turn out benign, some early tumours might still be missed.
Low-dose CT is not the same as a regular check-up, it is intended for people at higher risk, not for the general population.
Screening is only one part of the fight: quitting smoking, healthy lifestyle, regular medical check-ups, all remain important in reducing cancer risk overall.
***
The use of Sybil in Croatia shows how AI can move from research labs into everyday clinical practice, not as a pilot, but as a fully-fledged nationwide screening programme.
Using an AI tool like Sybil to analyze such a large volume of LDCT scans makes nationwide screening efficient, scalable and accessible. Early results already show that a substantial number of lung-cancer cases are caught in early, operable stages, offering many patients a real chance of cure, instead of facing cancer only after symptoms emerge.
