RESUMO
Vertebral body enhancement is occasionally seen on postcontrast CT imaging in the absence of osseous pathology. This enhancement can mimic sclerotic osseous metastatic disease, leading to a diagnostic dilemma for radiologists and increasing the chance of misinterpretation. Existing literature has focused on the association between this enhancement and concomitant central venous system obstruction. We report a 61-year-old woman with a history of nasopharyngeal carcinoma presenting with an epidural abscess who exhibited vertebral body enhancement resembling sclerotic metastatic disease without imaging evidence of central venous obstruction or vertebral osseous metastatic disease. Awareness of this unique presentation may prevent the incorrect diagnostic errors and their associated negative effects on patients.
RESUMO
Bone marrow edema-like lesions (BMEL) in the knee have been linked to the symptoms and progression of osteoarthritis (OA), a highly prevalent disease with profound public health implications. Manual and semi-automatic segmentations of BMELs in magnetic resonance images (MRI) have been used to quantify the significance of BMELs. However, their utilization is hampered by the labor-intensive and time-consuming nature of the process as well as by annotator bias, especially since BMELs exhibit various sizes and irregular shapes with diffuse signal that lead to poor intra- and inter-rater reliability. In this study, we propose a novel unsupervised method for fully automated segmentation of BMELs that leverages conditional diffusion models, multiple MRI sequences that have different contrast of BMELs, and anomaly detection that do not rely on costly and error-prone annotations. We also analyze BMEL segmentation annotations from multiple experts, reporting intra-/inter-rater variability and setting better benchmarks for BMEL segmentation performance.