Diffusion kurtosis imaging in predicting pathological grade of intrahepatic cholangiocarcinoma: a preliminary study

To investigate the diagnostic accuracy of diffusion kurtosis imaging (DKI) in predicting pathological grade of intrahepatic cholangiocarcinoma (ICC) preoperatively.

The study included 48 patients with ICC confirmed by histopathology (18 males, 30 females). The cohort comprised 24 moderately differentiated (G2, intermediate-grade) and 24 poorly differentiated (G3, high-grade) tumors. Preoperative MR examinations including DKI and diffusion-weighted imaging (DWI) were performed, and mean kurtosis (MK), mean diffusivity (MD), and apparent diffusion coefficient (ADC) values were measured. Diagnostic performance for predicting high-grade ICC via the receiver operating characteristic curve (ROC) was assessed. Logistic regression analyses were applied to assess the value of clinical and laboratory variables and diffusion parameters as risk factors for pathological grade.

High-grade ICC demonstrated significantly higher MK (0.85 ± 0.15 vs. 0.59 ± 0.12) but lower MD (1.38 ± 0.48 vs. 2.15 ± 0.76 × 10⁻³ mm²/s) and ADC (0.98 ± 0.23 vs. 1.36 ± 0.39 × 10⁻³ mm²/s) than intermediate-grade tumors (all P < 0.001). MK demonstrated superior diagnostic performance (area under ROC curve (AUC) = 0.925) compared to MD (AUC = 0.818) and ADC (AUC = 0.806) (all P < 0.001). Univariate analysis identified tumor size, MK, MD and ADC as predictors. Multivariate analysis confirmed MK (OR: 3.321; 95% CI: 1.559-10.37; P < 0.001) as an independent predictor.

DKI-derived MK outperforms MD and DWI-based ADC in predicting high-grade ICC, and could serve as a non-invasive preoperative biomarker to guide clinical management.