Purpose: To compare the accuracy of different MR sequences in simultaneous PET/MR imaging for T staging in non-small-cell lung cancer in relation to histopathology.
Methods: The study included 28 patients who underwent dedicated thoracic PET/MR imaging before tumour resection. Local tumour staging was performed separately by three readers with each of the following MR sequences together with PET: transverse T2 BLADE, transverse non-enhanced and contrast-enhanced T1 FLASH, T1 3D Dixon VIBE in transverse and coronal orientation, coronal T2 HASTE, and coronal TrueFISP. The staging results were compared with histopathology after resection as the reference standard. Differences in the accuracy of T staging among the MR sequences were evaluated using McNemar's test. Due to multiple testing, Bonferroni correction was applied to prevent accumulation of α errors; p < 0.0024 was considered statistically significant.
Results: Compared with histopathology, T-staging accuracy was 69% with T2 BLADE, 68% with T2 HASTE, 59% with contrast-enhanced T1 FLASH, 57% with TrueFISP, 50 % with non-enhanced T1 FLASH, and 45% and 48% with T1 3D Dixon VIBE in transverse and coronal orientation, respectively. Staging accuracy with T2 BLADE was significantly higher than with non-enhanced T1 FLASH and with T1 3D Dixon VIBE in transverse and coronal orientations (p < 0.0024). T2 HASTE had a significantly higher T-staging accuracy than transverse T1 3D-Dixon-VIBE (p < 0.0024).
Conclusion: Transverse T2 BLADE images provide the highest accuracy for local tumour staging and should therefore be included in dedicated thoracic PET/MR protocols. As T1 3D Dixon VIBE images acquired for attenuation correction performed significantly worse, this sequence cannot be considered sufficiently accurate for local tumour staging in the thorax.