AI’s prowess predicts lung cancer risk in non-smokers through routine chest X-rays. The study unveils an advanced “CXR-Lung-Risk” model, identifying 28% of participants at significantly higher risk. Dr. Lu emphasizes its pivotal role amid declining smoking rates.
Recent research presented at the Radiological Society of North America (RSNA) has unveiled a groundbreaking use of artificial intelligence (AI) in identifying non-smokers who face a high risk of developing lung cancer. This study introduces a novel approach that leverages routine chest X-ray images to predict the risk of lung cancer, especially in individuals who have never smoked or have minimal smoking history.
Lung cancer stands as the leading cause of cancer-related mortality, with a staggering number of new cases and deaths reported annually. Contrary to common perception, around 10-20% of lung cancer cases occur in individuals categorized as “never-smokers,” those who have either never smoked or have a minimal smoking history, involving fewer than 100 cigarettes throughout their lifetime.
Presently, prevailing guidelines for lung cancer screenings primarily target individuals aged 50 to 80 with an extensive smoking history of at least 20 pack-years, including current smokers or those who quit within the past 15 years. Regrettably, these recommendations often exclude never-smokers from screenings, leading to undetected cases where lung cancer manifests at later stages, posing graver challenges for effective treatment.
The study’s lead author, Anika S. Walia, highlighted the discrepancy between existing guidelines and the rising incidence of lung cancer among never-smokers. Walia emphasized, “Current guidelines limit lung cancer screening to individuals with significant smoking backgrounds. However, the surge in lung cancer cases among never-smokers often leads to diagnoses in advanced stages.”
The main hindrance in including never-smokers in screening recommendations lies in the complexity of predicting their lung cancer risk. Existing risk assessment tools necessitate data not routinely available for most individuals, such as detailed family history of lung cancer, pulmonary function tests, or specific biomarkers in the blood.
To bridge this gap, researchers at the Cardiovascular Imaging Research Center (CIRC) developed the “CXR-Lung-Risk” model. This sophisticated AI model was trained using 147,497 chest X-rays from both smokers and never-smokers, aiming to predict the risk of lung-related mortality solely based on a single chest X-ray image, a common medical test accessible in electronic medical records.
The model underwent validation in a separate cohort of never-smokers who underwent routine outpatient chest X-rays. The findings revealed that the high-risk group identified by the AI model, constituting 28% of the participants, demonstrated a significantly higher incidence of lung cancer, exceeding the recommended screening threshold.
Even after adjustments for various factors like age, sex, race, prior respiratory infections, and chronic obstructive pulmonary disease, the high-risk group identified by the AI model exhibited a 2.1 times greater risk of developing lung cancer compared to the low-risk group.
Dr. Michael T. Lu, the senior author of the study, highlighted the potential of this AI tool in facilitating opportunistic screenings for high-risk never-smokers using existing chest X-ray data in electronic medical records. Dr. Lu emphasized, “Given the decline in smoking rates, early detection strategies for lung cancer among non-smokers are becoming increasingly critical.”
The study benefited from support from various institutions, including the Boston University School of Medicine, the National Academy of Medicine/Johnson & Johnson Innovation Quickfire Challenge, and the Risk Management Corporation of the Harvard Medical Institutions Incorporated.
This innovative AI-enabled approach using readily available chest X-rays holds promising implications for early detection and intervention strategies tailored to non-smokers at high risk of lung cancer, potentially reshaping the landscape of lung cancer screening guidelines.
The AI-driven “CXR-Lung-Risk” model heralds a paradigm shift. Identifying high-risk individuals among non-smokers using chest X-rays, challenges conventional screening norms. Dr. Lu underscores the imperative need for expanded screenings in light of decreasing smoking rates. This innovation promises early interventions and potentially reshapes lung cancer screening guidelines.
