Differentiation between benign and malignant vertebral compression fractures using qualitative and quantitative analysis of a single fast spin echo T2-weighted Dixon sequence

Objectives To determine and compare the qualitative and quantitative diagnostic performance of a single sagittal fast spin echo (FSE) T2-weighted Dixon sequence in differentiating benign and malignant vertebral compression fractures (VCF), using multiple readers and different quantitative methods. Methods From July 2014 to June 2020, 95 consecutive patients with spine MRI performed prior to cementoplasty for acute VCFs were retrospectively included. VCFs were categorized as benign (n = 63, mean age = 76 +/- 12 years) or malignant (n = 32, mean age = 63 +/- 12 years) with a best valuable comparator as a reference. Qualitative analysis was independently performed by four radiologists by categorizing each VCF as either benign or malignant using only the image sets provided by FSE T2-weighted Dixon sequences. Quantitative analysis was performed using two different regions of interest (ROI1-2) and three methods (signal drop, fat fraction (FF) from ROIs, FF maps). Diagnostic performance was compared using ROC curves analyses. Interobserver agreement was assessed using kappa statistics and intraclass correlation coefficients (ICC). Results The qualitative diagnostic performance ranged from area under the curve (AUC) = 0.97 (95% CI: 0.91-1.00) to AUC = 0.99 (95% CI: 0.95-1.0). The quantitative diagnostic performance ranged from AUC = 0.82 (95% CI: 0.73-0.89) to AUC = 0.97 (95% CI: 0.91-0.99). Pairwise comparisons showed no statistical difference in diagnostic performance (all p > 0.0013, Bonferroni-corrected p < 0.0011). All five cases with disagreement among the readers were correctly diagnosed at quantitative analysis using ROI2. Interobserver agreement was excellent for both qualitative and quantitative analyses. Conclusions A single FSE T2-weighted Dixon sequence can be used to differentiate benign and malignant VCF with high diagnostic performance using both qualitative and quantitative analyses, which can provide complementary information.