Multiparametric analysis of monoexponential, intravoxelincoherent motion, and diffusion kurtosis models for differentiation of benignand malignant musculoskeletal tumors

Arslan Sarikaya S, Multiparametric analysis of monoexponential diffusion-weighted imaging, intravoxel incoherent motion, and diffusion kurtosis models for differentiation of benign and malignant musculoskeletal tumors, Hacettepe University, Faculty of Medicine, Department of Radiology, Thesis In Radiology, Ankara, 2021. The aim of our study was to evaluate the value of multiparametric analysis of monoexponential DWI, IVIM and DK models for differentiating benign and malignant musculoskeletal tumors. Between Aug 2019 to Nov 2020, a total of 50 patients (mean age: 32.8 years, range: 6-82 years) with benign and malignant musculoskeletal tumors were included prospectively. They were 28 cases of malignant tumors (13 female, 15 male patients, mean age: 30.36 years, range: 5-70 years) and 22 cases of benign tumors (12 female, 10 male, mean age: 35.90 years,range: 7-82 years) (bone tumors benign/malignant: 8/19, soft tissue tumors benign/malignant: 14/9). For all tumors, volumetric and minimum values of DWI (ADC), IVIM (D, D*,f), and DK (D, K) parameters were measured and compared between all benign and malignant tumors. These parameters were evaluated seperately for non-myxoid non-condroid (benign/malignant: 12/20) and myxoid and chondroid tumors (benign/malignant: 10/8). In all tumor group, D*vol, Kvol, Kmin, ADCmin, and DIVIM-min values showed statistically significant differences between malignant and benign tumors (P= 0.045, 0.002, 0.001, 0.026 and 0.015, respectively). D*vol, Kvol, and Kmin values were significantly higher, ADCmin and DIVIM-min values were significantly lower in malignant tumors compared with benign tumors. ROC analysis demonstrated that D*vol had the highest sensitivity with 75% and Kmin had the highest diagnostic performance with AUC of 0.760. In non-myxoid non-chondroid tumor group, D*vol, Kvol, Kmin, ADCmin, DIVIM-min, and Dkurtozis-min values showed statistically significant differences between malignant and benign tumors (P= 0.013, 0.035, 0.002, 0.005, 0.002, and 0.025, respectively). D*vol, Kvol, and Kmin values were significantly higher, ADCmin, DIVIM-min, and Dkurtozis-min values were significantly lower in malignant tumors compared with benign tumors. ROC analysis demonstrated that D*vol and DIVIM-min had the highest sensitivity with 95%, and Kmin had the highest diagnostic performance with AUC of 0.825. The cut-off values of D*vol, Kvol, Kmin, ADCmin, DIVIM-min, and Dkurtosis-min for discrimination of benign and malignant tumors in all tumor group and non-myxoid non-chondroid tumor group were as follows, respectively: 7.07/7.033x10-3 mm2/sn, 0.725/0.719,0.931/0.793, 1.027/1.027x10-3 mm2/sn, 0.999/0.999x10-3 mm2/sn, 1.132/1.132x10-3 mm2/sn. There was no significant differences in DWI, IVIM, and DK parameters to differentiate between benign and malignant myxoid chondroid tumors. In multivariate logistic regression analysis, no significant difference was detected between the diagnostic performance of DWI, IVIM, and DK in differentiating benign and malignant tumors. Three prediction models were designed, which demonstrated higher diagnostic performance than Kmin: 1st model: Kmin+DIVIM-min, 2nd model: Kmin+ADCmin, 3rd model: Kmin+ADCmin+DIVIM-min. ROC analyses of three logistic regression models showed that the ROC AUC of predicted probability increased by adding parameters to Kmin (AUC of 1st model: 0.783, 2nd model: 0.801, 3rd model: 0.801) without statistically significant differences (P >0.05). Keywords: Musculoskeletal tumors; Diffusion-weighted imaging; Intravoxel incoherent motion; Diffusion kurtosis; Multiparametric