Utility of a diffusion kurtosis model in the differential diagnosis of orofacial tumours

To compare the goodness of fit and correlations between diffusion kurtosis imaging (DKI) and a mono-exponential (ME) model, to compare the corrected apparent diffusion coefficient (D_app) and apparent kurtosis (K_app) of the DKI model, and the apparent diffusion coefficient (ADC) of the ME model among the various orofacial lesions, and to evaluate the diagnostic performances between the two models.

A total of 100 orofacial lesions underwent echo-planar diffusion magnetic resonance imaging (MRI) with four b-values. The goodness of fit was evaluated using Akaike information criterion. The correlations of the diffusion-derived parameters were evaluated. The diagnostic performance was analysed by receiver operating characteristics (ROC).

The DKI model showed a significantly better goodness of fit than the ME model (p<0.0001). The K_app had a strongly negative correlation with the D_app (ρ=-0.749) and ADC (ρ=-0.938). A strongly positive correlation existed between the D_app and ADC (ρ=0.906). All parameters differed significantly between benign tumours and malignant tumours (p<0.05). In differentiating benign tumours from the malignant tumours, the AUC of D_app (0.871) was larger than that of ADC (0.805); however, a significant difference was not found (p=0.102).

The DKI model had better goodness of fit than the ME model. Furthermore, the D_app and K_app were also characteristic for each pathological category; however, the DKI model did not yield a significantly higher diagnostic performance than the ME model, which might be related to the high correlation among the diffusion-derived parameters and wide variation among categories.