Imaging-based subtypes of pancreatic cancer
The outcomes of pancreatic cancer have seen little improvement over the past three decades. The disease is associated with a poor prognosis and lack of response to chemotherapy and radiation. To date, there are no reliable ways to separate patients into prognostic groups based on the cancer biology, which would be expected to help tailor therapy for each individual. To address the need to stratify patients, we hypothesized that diagnostic computed tomography imaging would reveal differences in the biological properties of pancreatic cancer and patient outcomes. We developed quantitative methods to analyze the CT images and diagnostic pathology of patients, associating these measurements with outcomes. We identified a distinct imaging signature of pancreatic cancer that associated with the amount of stroma in the tumor microenvironment, the biology of the cancer cells, response to treatments, and survival outcomes of the patients. Our findings indicate that quantitative characterization of routine diagnostic scans for pancreatic cancer can provide clinically and biologically meaningful information to help physicians make treatment decisions.