May 20, 2019 — Digital Reasoning announced results from its automated radiology report analytics research. In a series of experiments on radiology reports from emergency departments, inpatient and outpatient healthcare facilities, Digital Reasoning used natural language processing (NLP) and machine learning (ML) to identify and triage high-risk lung nodules, achieving queue precision of 90.2 percent. The findings have now been published in the Journal of Clinical Oncology as part of the 2019 American Society of Clinical Oncology (ASCO) meeting proceedings.1
For health systems, reviewing incidental findings can be a time and labor intensive process.2 Other studies show the rate for timely clinical follow-up can fall as low as 29 percent across the industry.3 Applying advanced artificial intelligence (AI) to radiology reports to automate the identification and triage of pulmonary nodules, empowers doctors to focus on reviewing and acting on the most high-risk cases. This results in improved patient safety and faster time-to-treatment without excess labor.
During the research study, Digital Reasoning analyzed 8,879 free-text, narrative computed tomography (CT) radiology reports from Dec. 8, 2015 through April 23, 2017. Today, those analytics are embedded in an enterprise solution utilized across more than 150 hospitals and 60 cancer centers in the United States.
For more information: www.ascopubs.org/journal/jco
1. French C., Makowski M., Terker S., et al. Automating incidental findings in radiology reports using natural language processing and machine learning to identify and classify pulmonary nodules. Presented at ASCO 2019. J Clin Oncol 37, 2019 (suppl; abstr e18093)
2. Rosenkrantz A.B., Xue X., Gyftopoulos S., et al. Downstream Costs Associated with Incidental Pulmonary Nodules Detected on CT. Acad Radiol., published online Aug. 6, 2018. pii: S1076-6332(18)30372-6, 2018