Recent studies examining the use of AI tools in lung cancer and radiology found that different tools garnered mixed results.
One study involved a retrospective analysis of over 25,000 x-rays taken between October 2025 and February 2026. Each x-ray was first reviewed by academic professionals and then by AI. The research team then compared the radiologists' reports with AI findings. They followed up with the patient if there was a discrepancy between the two reports.
The study showed that AI was good at prompting radiologists to double-check an x-ray, but it was less adept at identifying nodules on its own.
Another study analyzed the results of an AI tool that performed successfully in a lab setting and used in clinics to spot pulmonary nodules in CT scans. Bruno Hochhegger, the study team's leader, said that "the performance of this AI model decreased a lot" when applied to more than 10,000 patients.
Part of this is down to what data the AI was trained on. Additionally, if you train a tool for an outpatient center, for instance, it might work differently than a tool trained on emergency radiology patients.