The penumbra sign on T1-weighted MRI for differentiating musculoskeletal infection from tumour

Objective The penumbra sign on unenhanced T1-weighted MR sequences is thought to be helpful for discriminating subacute osteomyelitis from bone neoplasm. We sought to quantify the sensitivity and specificity of this sign for bone and soft tissue infection in a general referral population. Design Clinical coding was used to identify patients admitted to Middlemore Hospital (Auckland, New Zealand) between January 2000 and November 2003 with a diagnosis of either infection or neoplasm of the upper or lower limb who had undergone an MRI scan. One hundred and eighty-three patients were included in the study. Fifty-seven patients had bone or soft tissue infection. One hundred and twenty-six had a bone or soft tissue neoplasm. Relevant unenhanced T1-weighted images were selected for each patient, randomised and placed in a folder on the Hospital PACS system. Four reviewers were shown the original article describing the penumbra sign and then asked to look at the images in the folder stating whether the penumbra sign was present or absent. Results The average specificity and sensitivity of the penumbra sign for musculoskeletal infection was 96% (range 94-99%) and 27% (range 21-34%) respectively. Interobserver reliability was moderate to good with an average kappa score of 0.57 (range 0.50-0.62). For isolated soft tissue infection there was a higher sensitivity (54%, 33-83%) but with similar specificity (98%, 96-100%) and interobserver reliability (0.47, 0.33-0.58). Of the 11 out of 57 (19%) true positive penumbra signs identified by a consensus of three or more of the observers, all were subacute, chronic or acute on chronic infections. Conclusions The penumbra sign has a high specificity for musculoskeletal infection. This is also true for isolated soft tissue infection. The penumbra sign is helpful in differentiating neoplasm from infection and its presence in the setting of a high pretest probability is useful in making a diagnosis of infection.