ASO Author Reflections: Natural Language Processing Aids in the Detection of Incidental Pancreatic Lesions in Cross-Sectional Imaging

Excerpt

Pancreatic cancer continues to have a poor prognosis and is the third leading cause of cancer-related death in the United States.¹ While there have been advances in surgical technique and chemotherapeutic options, there is still significant morbidity and mortality associated with the disease. Therefore, a focus on maximizing the prevention of risk factors associated with pancreatic cancer is warranted. With the advent of advanced imaging technology over the last decade, incidental precancerous lesions such as intraductal papillary mucinous neoplasms (IPMNs) and pancreatic cysts have been identified at a greater rate. This provides a window of opportunity for earlier intervention in patients at risk for developing a pancreatic malignancy.² Prior studies have described the risk of malignancy for branch-duct IPMNs in the range of 12–17%, and 38–68% for main-duct IPMNs.³ Accordingly, early diagnosis and continued surveillance imaging for these lesions could potentially prevent future malignant transformation. Pancreatic incidentalomas on computed tomography imaging have been studied at length. Recent guidelines from the American College of Radiology recommended algorithms to help guide surveillance for patients with these high-risk lesions.³ Traditionally, these incidental findings have been identified through a manual review of imaging reports. More recently, studies have demonstrated the use of natural language processing (NLP) to expedite chart review of free-text reports, including imaging results.⁴ …