Different Attenuation Models of Diffusion‐Weighted MR Imaging for the Differentiation of Benign and Malignant Musculoskeletal Tumors.

Background: Several functional imaging techniques, including monoexponential diffusion‐weighted imaging (m‐DWI), intravoxel incoherent motion (IVIM), and diffusion kurtosis (DK) imaging, have been used in differentiating benign and malignant musculoskeletal tumors. Combining all three techniques in the same study population may improve differentiation. Purpose To compare the diagnostic performance of m‐DWI, IVIM, and DK models and their combinations in differentiating benign and malignant musculoskeletal tumors. Study Type: Prospective. Population: Fifty patients with benign and malignant musculoskeletal tumors divided into nonmyxoid and nonchondroid and myxoid and/or chondroid subgroups. Field strength/Sequence: A 1.5 T/m‐DWI, IVIM, and DK single‐shot spin‐echo echo‐planar sequences. Assessment Minimum and volumetric values of apparent diffusion coefficient (ADC), pure molecular diffusion (Divim), pseudodiffusion (D*), perfusion fraction (f), diffusion coefficient for kurtosis model (DK), and Kurtosis (K) were compared between all benign and malignant tumors. Subgroup analysis was also performed for nonmyxoid and nonchondroid and myxoid and/or chondroid tumors. Statistical tests: Independent samples t‐test, Mann–Whitney U test, intraclass correlation coefficient, ROC analysis, and logistic regression analysis. A P value < 0.05 was considered statistically significant. Results: ADCmin, Divim‐min, D*vol, DK‐min, Kvol, and Kmin values showed statistically significant differences between all benign and malignant tumors and nonmyxoid and nonchondroid tumor subgroup. Kmin showed the highest diagnostic performance in differentiating benign and malignant tumors with AUCs of 0.760 for "all tumors" and 0.825 for the nonmyxoid and nonchondroid tumor subgroup. No significant differences were detected in m‐DWI‐, IVIM‐, and DK‐derived parameters for differentiating benign and malignant myxoid and/or chondroid tumors. Only three of 63 combinations of prediction models demonstrated a higher diagnostic performance than Kmin; however, improvements were not significantly different. Data conclusion: ADCmin, Divim‐min, D*vol, DK‐min, Kvol, and Kmin values can be used to differentiate benign and malignant musculoskeletal tumors. Our findings suggest that the added value of multiparametric approach in such differentiation is not significant. Evidence Level 1 Technical Efficacy: Stage 2 [ABSTRACT FROM AUTHOR]