Gastroduodenal ulcers on CT: forgotten, but not gone.
OBJECTIVE: To estimate the incidence of missed gastroduodenal ulcers on routine abdominal computed tomography (CT) and identify findings and methods to improve sensitivity of CT interpretation for peptic ulcers. MATERIALS AND METHODS: This is a retrospective chart and imaging review. Two blinded readers independently reviewed CTs performed within 7 days prior to endoscopy of 114 subjects; this included 57 consecutive subjects with proven gastroduodenal ulcers intermixed with 57 subjects with endoscopically normal examinations. Presence, location and size of ulcer crater, and ancillary findings (mural edema, asymmetric wall thickening, focal fat stranding, regional lymph nodes, and extraluminal gas) were recorded before and after review of multiplanar reformatted images. Radiology reports were then reviewed to determine if an ulcer was identified prospectively. RESULTS: Thirty-one ulcers (54%) were radiographically occult, missed by both readers. Thirteen ulcers were correctly and independently identified by both readers (sensitivity/specificity = 30%/100%). With review of multiplanar reformats, sensitivity and accuracy increased for both readers. When two or more ancillary findings were identified, the odds ratio of a true ulcer being present was greater than 5.6 (P = 0.0001). Both size and location of ulcer were important for detection; readers were more likely to identify gastric ulcers compared to duodenal or marginal ulcers (P = 0.02). Only 3/13 definitely visible ulcers were correctly identified during initial CT interpretation. CONCLUSIONS: Although CT has low sensitivity for peptic ulcer disease, the miss rate for visible peptic ulcers is high. Increased awareness, multiplanar imaging review, and identification of ancillary findings may improve sensitivity for gastroduodenal ulcers.
Allen, BC; Tirman, P; Tobben, JP; Evans, JA; Leyendecker, JR
Volume / Issue
Start / End Page
Electronic International Standard Serial Number (EISSN)
Digital Object Identifier (DOI)