Purpose: Vertebral fractures are a predictor of further fractures and are associated with morbidity and mortality. The majority do not come to medical attention and are under-reported by radiologists. The Zebra Medical Vision Ltd. compression fracture detection algorithm can be applied to any computed tomography (CT) scan of the chest or abdomen. The aim was to evaluate the utility of the algorithm for the identification of vertebral fractures in patients having chest and/or abdomen CT scans performed for any reason at a tertiary referral hospital.
Methods: CT examinations of patients aged ≥50 years (104 abdominal; 103 chest) requested in the course of their management were evaluated. Acquisition was of consecutive studies over a two-week period. The Zebra algorithm was applied retrospectively to the digitised data. A manual search of the reports of these scans was undertaken to assess if a fracture had been noted, as well as an automated search using Report Analytics software. Applying standard criteria for the classification of vertebral fractures, scans were also assessed by an expert radiologist blind to the report and algorithm findings.
Results: The prevalence of vertebral fractures by expert radiology opinion was 16.5%. The sensitivity of the Zebra algorithm was 59% with specificity 94%, positive predictive value 67% and negative predictive value 92%. The sensitivity of routine reporting was low at 38%. Report Analytics identified 11 of 14 documented vertebral fractures.
Conclusion: The Zebra algorithm showed high specificity and compared to routine reporting, higher sensitivity. It offers an advance in the opportunistic detection of prevalent vertebral fractures.