This study evaluated an AI-assisted tool that helps doctors estimate the risk of cancer in lung nodules. The research team analyzed data from 951 patients who had these nodules surgically removed. The tool used a mathematical model that combined automatically extracted imaging features with clinical data and inflammatory markers to predict whether the nodule was malignant.
The analysis showed the tool had good accuracy, with a discrimination score of 0.836. When grouped by risk, the tool identified that lower-risk nodules had a 43.3% chance of being cancer, while higher-risk nodules had a 95.0% chance. The model performed well in this specific group of patients who were already undergoing surgery.
It is important to note that this study was limited to a group of patients selected specifically for surgery. The researchers warn that these results should not be used for screening or for nodules found incidentally in people who are not having surgery. Further testing in broader, independent groups of patients is required before this tool can be widely used outside of surgical settings.