Prediction of pancreatic neuroendocrine tumour grade with MR imaging features: added value of diffusion-weighted imaging

To evaluate the value of MR imaging including diffusion-weighted imaging (DWI) for the grading of pancreatic neuroendocrine tumours (pNET). Between 2006 and 2014, all resected pNETs with preoperative MR imaging including DWI were included. Tumour grading was based on the 2010 WHO classification. MR imaging features included size, T1-w, and T2-w signal intensity, enhancement pattern, apparent (ADC) and true diffusion (D) coefficients. One hundred and eight pNETs (mean 40 +/- 33 mm) were evaluated in 94 patients (48 women, 51 %, mean age 52 +/- 12). Fifty-five (51 %), 42 (39 %), and 11 (10 %) tumours were given the following grades (G): G1, G2, and G3. Mean ADC and D values were significantly lower as grade increased (ADC: 2.13 +/- 0.70, 1.78 +/- 0.72, and 0.86 +/- 0.22 10(-3) mm(2)/s, and D: 1.92 +/- 0.70, 1.75 +/- 0.74, and 0.82 +/- 0.19 10(-3) mm(2)/s G1, G2, and G3, all p < 0.001). A higher grade was associated with larger sized tumours (p < 0.001). The AUROC of ADC and D to differentiate G3 and G1-2 were 0.96 +/- 0.02 and 0.95 +/- 0.02. Optimal cut-off values for the identification of G3 were 1.19 10(-3) mm(2)/s for ADC (sensitivity 100 %, specificity 92 %) and 1.04 10(-3) mm(2)/s for D (sensitivity 82 %, specificity 92 %). Morphological/functional MRI features of pNETS depend on tumour grade. DWI is useful for the identification of high-grade tumours. Morphological and functional MRI features of pNETs depend on tumour grade. Their combination has a high predictive value for grade. All pNETs should be explored by MR imaging including DWI. DWI is helpful for identification of high-grade and poorly-differentiated tumours.