A University of Pittsburgh researcher has developed a novel computer tool to selectively identify patients at increased risk of lung cancer needing annual screening, using low dose computed tomography (LDCT).
Description
Screening for lung cancer, the leading cause of cancer deaths in the US, is carried out annually using LDCT. However, this is not without risk to healthy patients. This novel method will identify those most at risk of lung cancer and allows for individualization of lung cancer risk, stratification of that risk, and personalization of screening frequency. This approach has the potential to improve outcomes for patients, decrease the risk of unnecessary harm and reduce the economic burden of a wider screening program.
Applications
• Cancer screening
• Public health
• Harm reduction
Advantages
Currently large groups of people at an elevated risk for lung cancer are eligible for annual LDCT lung screening. This screening can reduce lung mortality by around 20% compared to chest X-rays, but LDCT is associated with a very high false positive rate resulting in unnecessary harmful follow up. Given most of those eligible for LDCT will never develop lung cancer, there is also risk of harm from unnecessary radiation exposure.
This novel approach determines lung cancer risk function over time based on LDCT scans from initial screening. Based on the risk function, a personalized screening frequency can be planned, in many cases reducing unnecessary annual screening. This approach will decrease the risk of harm from unnecessary LDCT scans and false positive results and has the potential to prevent lung cancer through the identification of early cancer signs before the appearance of suspicious lung nodules.
Invention Readiness
Analysis of data from the Pittsburgh Lung Screening Study (PLuSS) cohort focusing on factors derived from chest LDCT scans and patient demographics in combination with machine learning, led to the development of a model to determine lung cancer risk. Using historical scanning data to train the algorithm, clinical outcomes associated with a comparison cohort were used to demonstrate the results of the prediction tool. The automated tool was shown to predict the clinical outcome of patients in the PLuSS cohort, as well as identifying patients with a negligibly low risk of lung cancer, and those with a substantially higher than average risk.
IP Status
https://patents.google.com/patent/WO2024186858A1