Artificial intelligence has shown promise in identifying elevated cancer risk among individuals who have never smoked, using chest X-rays. Findings from this line of work will be highlighted at a major radiology conference in North America.
While smoking remains the strongest established risk factor for lung cancer, a meaningful minority—roughly 10% to 20% of cases—occur in people who have never smoked or who have smoked very few cigarettes in their lifetime.
In a recent study, researchers from a leading medical school developed the CXR-Lung-Risk deep learning model. The team trained the algorithm on a dataset of 147,497 chest X-ray images drawn from more than 40,000 individuals.
The model was evaluated on 17,407 participants with an average age around 63. The AI categorized about 28% as high risk for lung cancer. After adjusting for sex, age, and prior lung infections, the risk in the high-risk group remained 2.1 times greater than in the low-risk group. Among these high-risk individuals, 2.9% were later diagnosed with lung cancer. This incidence suggests that those identified as high risk should be considered for computed tomography (CT) screening, which provides more detailed information than a chest X-ray.
The study demonstrates that AI can help pinpoint never-smokers who may benefit from more comprehensive lung cancer screening protocols.
In related work, researchers have explored AI approaches to detect other cancers, such as stomach cancer, achieving high diagnostic accuracy in different datasets, underscoring the potential of artificial intelligence to augment cancer risk assessment and screening strategies across diverse patient groups.