Purpose: To compare fat fraction (FF) and signal drop from in-phase to opposed phase (I-O) derived from T1 VIBE and T2 turbo spin-echo (TSE) Dixon magnetic resonance imaging (MRI) for discrimination between malignant and healthy spinal marrow.
Methods and Materials: A total of 30 patients with vertebral metastases of known primary malignancy and 30 patients without a history of vertebral malignancy were included in this retrospective study. Each had received 1.5-T spinal MRI including T1 VIBE and T2 TSE two-point Dixon MRI with dedicated I-O echo timing. The FF was calculated by division of the fat-only through the sum of the water- and fat-only images. The I-O was calculated as in-phase minus opposed phase divided by in-phase images. Three-dimensional volumes (VOIs) comprising one vertebral metastasis or one healthy vertebra were generated on the T1 VIBE Dixon FF images and copied onto the T2 FF. The mean FF value was noted. Regions of interest (ROIs) were drawn within a representative slice of the VOI on T1 VIBE FF images and copied onto the T2 TSE FF as well as T1 and T2 in- and opposed phase images. Mean ROI signal intensity values were noted and I-O was calculated. Then t tests, area under the curve receiver operating characteristic (ROC AUC) analyses, and Pearson correlation coefficients (r) were used to compare and correlate mean measurement values, AUC, and determining optimized cutoff values for discrimination between malignant and benign vertebral marrow. A p value < 0.05 was deemed statistically significant.
Results: Malignant mean VOI T1 FF (11%), ROI T1 FF (9%), T2 VOI FF (9%), T2 ROI FF (7%), T1 I-O (13%), and T2 I-O (18%) were significantly smaller in malignant compared with healthy marrow (68%, 73%, 74%, 77%, 58%, and 51%, respectively; each p < 0.001). VOI T1 FF, ROI T1 FF, T2 VOI FF, and T2 ROI FF AUC were 1.00 (95% confidence interval [CI], 1.00–1.00). T1 and T2 I-O AUC were 0.998 (95% CI, 0.992–1) and 0.964 (95% CI, 0.90–1.00), respectively. VOI T1 FF, ROI T1 FF, T2 VOI FF, and T2 ROI FF showed 100% sensitivity and 100% specificity at cutoff values of 34%, 32%, 36%, and 34%, respectively. A T1 I-O cutoff of 16% and T2 I-O cutoff of 18% each showed 93% sensitivity and 100% specificity for identification of malignant vertebral marrow. ROC AUC did not differ significantly between the parameters (each p > 0.15). Moderate positive correlation was found between the FF measurements for malignant (each r > 0.65) and healthy vertebrae (each r > 0.54), and no significant correlation was found between FF and I-O measurements in the malignant cohort (each p > 0.12). There was no significant correlation between T1 and T2 I-O measurements in the malignant cohort (p = 0.06).
Conclusion: Quantitative T1 and T2 Dixon MRI parameters FF and I-O can reliably discriminate between malignant and healthy vertebral marrow. There was no significant difference in diagnostic performance between T1 and T2 Dixon MRI.
