skip to main content
Research Highlight | January 15, 2020

Lung study team optimistic about new algorithm that can detect lung cancer up to three years before it forms

TFRI’s Early Detection Lung Study team created a world-leading predictive model that has helped to save the lives of many lung cancer patients by detecting and treating it earlier. Now they have developed a way to help doctors predict who is at higher risk of developing the disease. Their deep learning algorithm (Deep LR) can detect the disease up to three years before it ever forms.

BC Cancer respirologist Dr. Stephen Lam and his research team shared their research results in The Lancet (October 2019).  The algorithm was created with the use of two large cohorts. The training cohort, which consisted of CT scans and corresponding outcomes of 25,097 patients, was first fed to the algorithm, which analyzed several nodule and non-nodule features to determine who was at a higher risk of developing lung cancer up to three years after a CT scan. These results were then validated using a second cohort of 2,350 patients, which was part of a TFRI-funded clinical trial (2008 Pan-Canadian Early Detection Lung Study).

According to Dr. Lam, the algorithm outperformed existing predictive lung cancer models, accurately stratifying high-risk patients from low-risk ones. These results mean that the algorithm could be used to guide the timing of surveillance lung cancer screening CT scans, increasing the chances of detecting aggressive lung cancer  while reducing the resource burden of health care systems through prolonging the interval of the next surveillance scan given to low-risk patients.

“DeepLR accounted for all relevant nodule and non-nodule features on screening chest CT scans and accurately predicted the presence of lung cancer within a three-year period,” says Dr. Lam. “Our study provides the framework to prospectively assess different screening intervals and more urgent diagnostic approaches for suspicious lung nodules based on malignancy risk.”


Prediction of lung cancer risk at follow-up screening with low-dose CT: a training and validation study of a deep learning method


Peng Huang, Cheng T Lin, Yuliang Li, Martin C Tammemagi, Malcolm V Brock, Sukhinder Atkar-Khattra, Yanxun Xu, Ping Hu, John R Mayo, Heidi Schmidt, Michel Gingras, Sergio Pasian, Lori Stewart, Scott Tsai, Jean M Seely, Daria Manos, Paul Burrowes, Rick Bhatia, Ming-Sound Tsao and Stephen Lam


The study was partially funded by the Terry Fox Translational Research Program grant to Early Detection of lung cancer