Researchers at MIT have developed an AI deep learning model called ‘Sybil’ that can predict the risk of lung cancer up to six years in advance through a single low-dose CT scan.
The current lung cancer prediction models require demographic information, clinical risk factors, and radiologic annotations, whereas Sybil only requires a single low-dose chest scan to predict the risk of lung cancer occurrence. The study showed it was able to forecast both short-term and long-term lung cancer risk with high accuracy.
The imaging data used to train Sybil was largely absent of any signs of cancer, as early-stage lung cancer occupies small portions of the lung. Despite the lack of visible cancer, it was able to predict which lung would eventually develop cancer, demonstrating its predictive power.
The researchers hope that this model will bring the research community one step closer to outgrowing legacy systems in the healthcare industry and help better treat current and future patients.
Sybil’s ability to predict lung cancer risk up to six years in advance through a single low-dose CT scan has the potential to revolutionize lung cancer screening and treatment. What do you think of this development?