Researchers studied whether an artificial intelligence (AI) model could help doctors predict if very small, hazy spots on lung CT scans are invasive cancer. They looked at data from 260 patients at two hospitals who had these small nodules and later had surgery to confirm their diagnosis. The AI model analyzed specific features from the CT scans to make its predictions.
In the first hospital's data, the model correctly identified invasive cancer about 79% of the time and correctly ruled it out about 73% of the time. When tested on data from a second hospital, its performance was slightly lower but still showed promise. The model used 10 different features from the CT images to make its calculations.
No safety issues were reported because this study only analyzed existing patient data; no new treatments were given. The main reason for caution is that this was a retrospective study, meaning it looked back at old patient records. This type of study design can sometimes overestimate how well a model will work in real-time clinical practice.
Readers should understand that this research represents an early step in developing better tools for lung cancer diagnosis. The AI model is not ready for routine use yet and needs to be tested in prospective studies where doctors use it to help make real-time decisions. If validated further, such tools might one day help doctors and patients decide when surgery is most appropriate for these challenging small lung nodules.