Project Co-Leads: Drs. Rayjean Hung and Geoffrey Liu
Vision: Analyzing CT images using Artificial Intelligence (AI) techniques will better distinguish cancerous from benign tissue.
Less than 10-15% of lesions seen on scans are actually lung cancer. Although we have been able to reduce the number of biopsies on false positive lesions, they are still performing more short-term scans to watch these spot closely. This leads to increased radiation exposure, needless patient anxiety and higher healthcare costs.
Artificial intelligence (AI) tools–known as radiomics–will be used to help distinguish between cancerous and benign spots seen on low dose computed tomography scans (LDCTs). These tools will extract and analyze large amounts of data from LDCTs. This mined data has the potential to uncover cancer characteristics that cannot be seen on scans by the naked eye. We hope that radiomics will help reduce uncertainty around whether masses seen in scans are, in fact, cancerous. As a result, fewer patients would need unnecessary follow-up procedures, and overall costs would be reduced.
We will analyze over 15,000 LDCTs from the Toronto I-ELCAP study. After, we will focus on three international studies–the U.S. National Lung Screening Trial, the Dutch NELSON trial, and the UK Liverpool trial–to determine if our AI tools offer high accuracy on their patients’ LDCTs.
What will the impact be five years from now?
Our robust team, with its experience in AI analysis and blood-based testing, will have the expertise and resources to differentiate not only if a mass is cancerous or not, but the type of lung cancer as well. This could have a huge impact on our ability to treat patients sooner, and may even replace the traditional invasive and expensive tissue biopsies.