According to a new study, a program working with artificial intelligence (AI) can examine if a non-smoker has a high lung cancer risk. The AI program does so by simply examining a single x-ray scan.
Lung Cancer Risk Among Non-Smokers
Dr. Michael T. Lu, an author of the study and the co-director of the Massachusetts General Hospital's Cardiovascular Imaging Research Center, explains that the model leads to opportunities to screen never-smokers and see if they have a high lung cancer risk through existing x-rays on the chest.
This is quite crucial, as the NCCN (National Comprehensive Cancer Network) only recommends individuals who have a 1.3% lung cancer risk or a high risk of getting lung cancer in the next six years undergo lung cancer CT scans. The latter risk group covers smokers or those who have a family history of lung cancer.
It is important to note that 10% to 20% of lung cancer cases involve non-smokers. However, doctors do not have any method of predicting high lung cancer risk among non-smokers. Because of this, they have been excluded from federal screening recommendations and guidelines.
ALSO READ : Air Pollution Triggers Lung Cancer in Non-Smokers by Increasing Gene Mutation Linked to the Disease
AI Model Assesses Non-Smoker Lung Cancer Risk
As part of the new study, the new AI model named CXR Lung-Risk was tested on thousands of non-smokers chest x-rays. The ages of the involved participants ranged from 55 to 74 years old.
The model was able to distinguish that 28% of the individuals had a high lung cancer risk. Findings were presented by the researchers during the RNSA's (Radiological Society of North America) annual meeting last November 26-30.
In the high-risk group, almost three out of 100 were found to develop lung cancer in the next six years. This is more than twice the minimum thresholds that lead to recommended screenings.
The model is a deep-learning one, which means that it learns through pattern recognition and from the own experience of the system in moving the information across various neural network layers. The algorithm was trained by researchers with 40,643 asymptomatic smokers' 147,497 chest x-rays. It was also trained with the x-rays of non-smokers from the PLCO (Prostate, Lung, Colorectal, and Ovarian) Cancer Screening Trial that took place from 1993 to 2001. Some of the individuals ended up developing lung cancer within the six years of being part of the trial.
The AI program also looked into x-rays from 2013 to 2014 that were not labeled. These were from 17,047 non-smokers. It did so to look into future diagnosis odds. It categorized patients into groups depending on risk levels.
In the high-risk group, the 2.9% patients that ended up getting lung cancer had a 2.1 times higher risk of getting lung cancer compared to the group with low risk. They also significantly went beyond the 1.3% threshold of risk for triggering a recommended screening.
Dr. Lu explains that as the rates of cigarette smoking are going down, it will be growingly important to detect lung cancer early among those who are not smokers.
RELATED ARTICLE : Lung Cancer Patients Who Quit Smoking Even After Diagnosis Have Higher Chances of Survival Rate, Study Says
Check out more news and information on Medicine & Health in Science Times.
