An Ontology-Based Approach to Estimate the Frequency of Rare Diseases in Narrative-Text Radiology Reports.

This study sought to use ontology-based knowledge to identify patients with rare diseases and to estimate the frequency of those diseases in a large database of radiology reports. Natural language processing methods were applied to 12,377,743 narrarive-text radiology reports of 7,803,811 patients at an academic health system. Using knowledge from the Orphanet Rare Disease Ontology and Radiology Gamuts Ontology, 1,154 of 6,794 rare diseases (17.0%) were observed in a total of 237,840 patients (3.05%). Ninety of 2,129 diseases (4%) with known prevalence less than 1 per 1,000,000 were observed in the database, whereas 100 of 173 diseases (58%) with prevalence greater than 1 per 10,000 were observed; the difference was statistically significant (p < .00001). Automated ontology-based search of radiology reports can estimate the frequency of rare diseases, and those diseases with higher known prevalence were significantly more likely to appear in radiology reports.