MRI-based anatomical landmarks for the identification of thoracic vertebral levels

Aim: To identify soft-tissue and bony anatomical landmarks on dedicated thoracic spine magnetic resonance imaging (MRI), and to assess their detectability, reproducibility, and accuracy in predicting specific thoracic vertebral levels. Materials and methods: One hundred dedicated thoracic MRI studies were retrospectively analysed by two radiologists independently. Ten bone and soft-tissue landmarks were localized to the adjacent vertebral level. The true numerical thoracic vertebral level was subsequently determined and recorded by cross referencing with a sagittal cervico-thoracic "counting scan". Results: Six landmarks were defined in ≥98% cases; however, there was a low interobserver percentage agreement for the defined vertebral levels (>70% for only one landmark). The most useful landmark for defining a specific vertebral level was the most superior rib (98% detection, 95% interobserver agreement, 98% at a single vertebral level, 0.07 SD). Eight landmarks localized to a specific thoracic segment in only 16–44% of cases, with a standard deviation of >0.5 vertebral levels and with a range which was greater than four vertebral levels. Conclusion: The C2 vertebra must be identified and cross referenced to the dedicated thoracic spine MRI, as other MRI-based anatomical landmarks are unreliable in determining the correct thoracic vertebral level