Risk of Lung Cancer predicted by using Single LDCT

A new deep learning model has demonstrated the ability to predict lung cancer risk using a single low-dose computed tomography (LDCT) screening exam.

Recent research presented at the American Thoracic Society 2025 International Conference highlighted the capabilities of this model, known as Sybil. Developed using imaging data from the National Lung Cancer Screening Trial (NLST), Sybil offers a promising approach to diagnosing lung cancer risk assessment. According to experts, Sybil could enhance risk stratification and help tailor screening strategies to individual patients.

The model was evaluated using over 21,000 LDCT scans from individuals aged 50 to 80 who voluntarily participated in screening between 2009 and 2021. Patient outcomes were tracked through 2024 to assess the model’s predictive accuracy.

Unlike conventional risk models, which rely on a range of clinical and demographic variables, Sybil was designed to predict future lung cancer development based solely on imaging data. Researchers found that Sybil performed strongly, effectively forecasting lung cancer risk at both one-year and six-year intervals.

Yeon Wook Kim, MD, a pulmonologist and researcher at Seoul National University Bundang Hospital in Seongnam, Republic of Korea, and colleagues noted “Sybil’s value lies in its unique ability to predict future lung cancer risk from a single LDCT scan, independent of other demographic factors that are conventionally used for risk stratification.” 

The team is optimistic that the model could be used to more accurately screen for patients who have the highest risk of developing cancer, allowing low-risk patients to go longer between screenings. 

“Sybil demonstrated the potential to identify true low-risk individuals who may benefit from discontinuing further screening, as well as to detect at-risk groups who should be encouraged to continue screening,” the group suggested. 

In areas such as Asia, experiencing increases in lung cancer in non-smokers, this tool could be especially beneficial.