Image: AI identification of lung nodule matches or bests that of trained radiologists (Photo courtesy of iStock)
Combining artificial intelligence (AI) and lung imaging reporting and data system (Lung-RADS) scores can increase CT scan specificity without reducing sensitivity, according to a new study.
Researchers at the University of Saskatchewan (Saskatoon, Canada) conducted a study that performed secondary analysis of a known data set using an AI model developed by Google in 2019, and Lung-RADS classifications from six radiologists. They then compared them to assess a representative cohort of 3,197 baseline low-dose CT screening patients. To ensure the AI algorithm matched the 91% sensitivity level achieved by the providers, the researchers determined a 0.27 AI risk-score threshold, based on a 0-to-1 scale.
The results showed that the AI-informed management strategy achieved sensitivity and specificity of 91% and 96%, respectively, while the average sensitivity and specificity of the six radiologists using only Lung-RADS was 91% and 61%, respectively. Based on the AI management strategy, 0.2% of category 1 or 2 Lung-RADS classifications were upgraded to category 3, and 30% of category 3 or higher classifications were downgraded to category 2. The minimum net cost savings, based on 2019 U.S. Medicare physician fee schedule, was USD 72 per patient screened. The study was published on January 19, 2021, in Journal of the American College of Radiology.
“Using an AI risk score combined with Lung-RADS at baseline lung cancer screening may result in fewer follow-up investigations and substantial cost savings. Specificity could rise by more than fifty percent,” concluded lead author Scott Adams, MD, and colleagues. “Additional research for other AI thresholds could also beneficial, especially for Lung-RADS category 4 nodules. Ultimately, additional investigations could lead to AI algorithms being used in a similar way to what has been suggested for screening mammography.”
Lung-RADS is a quality assurance tool designed to standardize lung cancer screening CT reporting and management recommendations, reduce confusion in lung cancer screening CT interpretations, and facilitate outcome monitoring. It is modeled on the success of the Breast Imaging Reporting and Data System (BI-RADS), with the primary goal of minimizing variation in the management of CT-detected lung nodules so that screening can be implemented effectively in radiology practices.
University of Saskatchewan