An artificial intelligence (AI) model may be capable of using routine chest x-ray images to identify nonsmokers who may be at high risk for lung cancer, according to new findings presented by Walia et al at the Radiological Society of North American (RSNA) 2023 Scientific Assembly and Annual Meeting (Abstract M5A-SPCH-6).
Background
Lung cancer is the most common cause of cancer mortality. The American Cancer Society estimated that there will be 238,340 new diagnoses and 127,070 deaths from lung cancer in 2023. Approximately 10% to 20% of lung cancer cases occur in never-smokers, those who have never smoked or smoked fewer than 100 cigarettes in their lifetime.
Deep learning is an advanced type of AI that can be trained to search x-ray images to find patterns associated with disease.
The United States Preventive Services Task Force (USPSTF) recommends that patients aged 50 to 80 years with at least a 20 pack-year smoking history who currently smoke or have quit smoking within the past 15 years should undergo lung cancer screening with low-dose computed tomography (CT). Although the USPSTF does not recommend screening for patients who have never smoked or who have smoked very little, the incidence of lung cancer among never-smokers is on the rise. Without early detection through screening, these types of cancers tend to be more advanced at diagnosis than those found in smokers.
“Current Medicare and USPSTF guidelines recommend lung cancer screening CT only for [patients] with a substantial smoking history," explained lead study author Anika S. Walia, BA, a medical student at the Boston University School of Medicine and a researcher at the Cardiovascular Imaging Research Center at Massachusetts General Hospital and Harvard Medical School. "However, lung cancer is increasingly common in never-smokers and often presents at an advanced stage,” she stressed.
A potential reason behind the exclusion of never-smokers from screening recommendations in federal guidelines may be because it is often difficult to predict lung cancer risk in this patient population. Existing lung cancer risk scores require information that is not readily available for most patients, such as a family history of lung cancer, pulmonary function testing, or serum biomarkers.
Study Methods and Results
In the new study, researchers used 147,497 chest x-rays from 40,643 asymptomatic smokers and never-smokers who participated in the Prostate, Lung, Colorectal, and Ovarian cancer screening trial to develop the novel CXR-Lung-Risk deep learning model and predict lung-related mortality risk based on a single chest x-ray image as input.
“A major advantage to our approach is that it only requires a single chest x-ray image, which is one of the most common tests in medicine and widely available in the electronic medical record," Ms. Walia emphasized.
The researchers then analyzed the efficacy of the CXR-Lung-Risk model by testing whether it could identify high-risk patients based on chest x-rays from their electronic medical records—with the goal of improving lung cancer risk prediction in never-smokers. Further, they externally validated the AI model in a separate group of never-smokers who underwent routine outpatient chest x-rays from 2013 to 2014.
The primary outcome was 6-year incident lung cancer, identified using International Classification of Disease codes. Risk scores were then converted to low-, moderate-, and high-risk groups based on externally derived risk thresholds.
The researchers discovered that among the 17,407 patients involved in the study, 28% of them were deemed high risk by the CXR-Lung-Risk model, and 2.9% of them later had a diagnosis of lung cancer. They reported that the high-risk group exceeded the 1.3% 6-year risk threshold where lung cancer screening CT is recommended by the National Comprehensive Cancer Network guidelines.
After adjusting for age, sex, race, previous lower respiratory tract infection, and prevalent chronic obstructive pulmonary disease, the researchers found that there was still a 2.1-times greater risk of developing lung cancer in the high-risk group compared with the low-risk group.
Conclusions
"This AI [model] opens the door for opportunistic screening for never-smokers at high risk of lung cancer, using existing chest x-rays in the electronic medical record," underscored senior study author Michael T. Lu, MD, MPH, Director of Artificial Intelligence and Co-Director of the Cardiovascular Imaging Research Center at Massachusetts General Hospital. "Since cigarette smoking rates are declining, approaches to detect lung cancer early in those who do not smoke are going to be increasingly important,” he concluded.
Disclosure: The research in this study was supported by the Boston University School of Medicine Student Committee on Medical School Affairs, the National Academy of Medicine/Johnson & Johnson Innovation Quickfire Challenge, and the Risk Management Corporation of the Harvard Medical Institutions Incorporated.
