Smoking is the leading cause of lung cancer. However, even among those who have never smoked, rates of lung cancer are steadily increasing. Often, the disease remains asymptomatic until it has spread to other organs, rendering traditional treatments ineffective. While scientists have successfully developed statistical models predicting lung cancer risk in smokers with no symptoms, the same success has not been achieved for never-smokers.
Recent advances in machine learning, employing computer applications capable of predicting patient outcomes without human pre-programming, hold promise for predicting lung cancer in people who have never smoked. This research aims to investigate whether machine learning surpasses traditional predictive models developed by scientists. Using demographic, lifestyle, environmental, genetic, and additional data from the UK Biobank, we will build investigator-based and machine learning-based models. Subsequently, we will assess the prediction accuracy of each model.
The ultimate goal of this research is to develop an efficient prediction model that informs future studies identifying high-risk populations among people who have never smoked. Such findings may contribute to the formulation of policies and strategies for lung cancer screening tests in this population. Early detection of lung cancer in people who have never smoked can significantly enhance treatment outcomes, patient survival, and quality of life, thereby easing the burden of treatment costs on both patients and the healthcare system.
The project duration is expected to span three years.
