At the forthcoming annual meeting of the Radiological Society of North America (RSNA), a study will be unveiled that presents an artificial intelligence (AI) solution capable of identifying non-smokers at a heightened risk of lung cancer through the analysis of standard chest X-ray images.
Lung cancer stands as the primary cause of cancer-related deaths, with 127,070 fatalities and approximately 238,340 new cases documented in the US this year, according to the American Cancer Society. It is noteworthy that 10 to 20% of lung tumors manifest in individuals classified as “never smokers,” indicating those who have either never engaged in smoking or have consumed fewer than 100 cigarettes throughout their lives.
For individuals aged between 50 and 80 with a smoking background of at least 20 pack years, the United States Preventive Services Task Force (USPSTF) advises lung cancer screening via low-dose CT scans for current smokers or those who ceased smoking within the last 15 years. Nevertheless, the incidence of lung cancer among non-smokers is escalating, often resulting in the detection of more advanced tumors due to delayed diagnosis.
Anika S. Walia, the principal investigator of the research and a medical student at Boston University School of Medicine and Harvard Medical School, underscored the current guidelines’ emphasis on individuals with substantial tobacco consumption history while overlooking the rising occurrence of lung cancer among non-smokers.
To bridge this disparity, scientists at the Cardiovascular Imaging Research Center (CIRC) formulated a sophisticated deep learning algorithm dubbed “CXR-Lung-Risis” to forecast the risk of lung cancer in non-smokers based on chest X-ray images extracted from electronic health records. This model, trained on an extensive dataset from the Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer screening trial, exhibited encouraging outcomes in pinpointing high-risk individuals.
The research scrutinized the model using a cohort of never-smokers who underwent regular chest X-rays, segregating participants into low, moderate, and high-risk categories predicated on their lung cancer risk scores. The high-risk cluster identified by the AI system displayed a notably elevated likelihood of developing lung cancer in comparison to the low-risk cluster, even after adjusting for variables such as age, sex, and medical background.
The investigators accentuated the capacity of this AI tool to facilitate tailored screening for never-smokers at an escalated risk of lung cancer, leveraging easily accessible chest X-ray images stored in electronic medical records. This breakthrough harbors substantial potential for early detection strategies, particularly in light of declining smoking prevalence.
The financial backing for this study was extended by the National Academy of Medicine/Johnson & Johnson Innovation Quickfire Challenge, the Risk Management Corporation of the Harvard Medical Institutions Incorporated, and the Boston University Student Committee on Medical School Affairs.
