July 25, 2024

Machine Learning Advances Simplify Lung Cancer Screening Process

Researchers from UCL and the University of Cambridge have developed machine learning models that can accurately predict an individual’s risk of developing lung cancer within five years. These models are as good or even better than the best risk models currently available, while using only a quarter of the information required. The study, published in PLOS Medicine, brings personalized lung cancer screening closer to reality.

Lung cancer is the leading cause of cancer death globally, with a low survival rate in the absence of early detection. It is estimated that in 2020, there were 1.8 million deaths due to lung cancer worldwide. Screening individuals at high risk for lung cancer could potentially reduce lung cancer-specific mortality by 20-24%, but determining who is high-risk and implementing effective screening approaches remains a challenge.

The UK is planning a national screening program for lung cancer, targeting individuals aged 55-74 who have smoked in their lifetime. Currently, the screening program relies on a risk model that requires answering 17 complex questions. This information is time-consuming to obtain and would require a call center with 50-100 employees to collect data from one million individuals.

In this study, researchers used data from the UK Biobank and the US National Lung Screening Trial to develop machine learning models that simplify the prediction of an individual’s risk of developing lung cancer within the next five years. By experimenting with over 60 different machine learning pipelines, the researchers identified the most effective models using only three variables: age, smoking duration, and average number of cigarettes smoked per day.

Four model pipelines were selected and combined into an ensemble model that accurately predicted lung cancer risk, using only a third of the variables required by existing models. This simplification greatly reduces the burden of data collection for screening programs. Dr. Tom Callender from UCL Medicine, the first author of the study, expressed the importance of this approach, stating, “Our study shows that artificial intelligence can be used to accurately predict lung cancer risk using just three pieces of information that would be easy to gather during routine GP appointments, online, or via apps. This approach has the potential to greatly simplify population level screening for lung cancer and help to make it a reality.”

The models used in the study were externally validated in the US Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial and compared against established models. The researchers believe this approach could also simplify the screening process for other diseases such as type-2 diabetes and cardiovascular disease.

The authors of the study hope that their findings will contribute to the development of a national lung cancer screening program that is quicker, more accessible, and more cost-effective, while still achieving the primary goal of reducing lung cancer mortality. Professor Sam Janes from UCL Medicine, the senior author of the study, emphasized the importance of feasibility and participation in any national screening program, stating, “Our findings are good news on both counts.”

This research was supported by Wellcome, the National Science Foundation, the Medical Research Council, and Cancer Research UK. With further advancements in machine learning and the potential for widespread implementation, lung cancer screening could become more efficient and successful in saving lives.

*Note:
1. Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it