A study on diffusion and kurtosis features of cervical cancer based on non-Gaussian diffusion weighted model

To explore the diffusion and kurtosis features of cervical cancer (CC) and study the feasibility of diffusion kurtosis imaging (DKI) based on the non-Gaussian diffusion-weighted model to differentiate the stage and grade of CC.

A total of 50 patients with pathologically confirmed CC were enrolled. MRI examinations including DKI (with 5b values 200, 500, 1000, 1500, and 2000smm^-2 were performed before any treatment. The apparent coefficient (D_app) and the apparent kurtosis value (K_app) were derived from the non-gaussian diffusion model, and the apparent diffusion coefficient (ADC) was derived from the Gaussian model. The parameters of CC and normal tissue (myometrium) were obtained, analyzed statistically, and evaluated with respect to differentiating stage and grade between the tissue and the CC.

ADC and D_app values of CC were significantly lower than that of the normal myometrium (P=0.024 and P<0.001, respectively), while the K_app value was not found to exhibit a significant difference. Compared to the well/moderately differentiated CC, poorly differentiated CC had a significantly decreased mean ADC and D_app (P=0.018 and P=0.026, respectively); however, the mean K_app (P=0.035) increased significantly. In the clinical staging, the DKI sequence was advantageous over conventional MRI sequences (degree of accuracy: 90% vs. 74%), Although in the quantitative analysis, these parameters did not show a significant difference.

The pilot study demonstrated that these diffusion and kurtosis indices from DKI based on the non-Gaussian diffusion-weighted model putatively differentiated the grade and stage of CC.