Diffusion-weighted MR imaging in primary rectal cancer staging demonstrates but does not characterise lymph nodes

Abstract

To evaluate the performance of diffusion-weighted MRI (DWI) for the detection of lymph nodes and for differentiating between benign and metastatic nodes during primary rectal cancer staging. Twenty-one patients underwent 1.5-T MRI followed by surgery (± preoperative 5 × 5 Gy). Imaging consisted of T2-weighted MRI, DWI (b0, 500, 1000), and 3DT1-weighted MRI with 1-mm isotropic voxels. The latter was used for accurate detection and per lesion histological validation of nodes. Two independent readers analysed the signal intensity on DWI and measured the mean apparent diffusion coefficient (ADC) for each node (ADCnode) and the ADC of each node relative to the mean tumour ADC (ADCrel). DWI detected 6 % more nodes than T2W-MRI. The signal on DWI was not accurate for the differentiation of metastatic nodes (AUC 0.45-0.50). Interobserver reproducibility for the nodal ADC measurements was excellent (ICC 0.93). Mean ADCnode was higher for benign than for malignant nodes (1.15 ± 0.24 vs. 1.04 ± 0.22 *10(-3) mm(2)/s), though not statistically significant (P = 0.10). Area under the ROC curve/sensitivity/specificity for the assessment of metastatic nodes were 0.64/67 %/60 % for ADCnode and 0.67/75 %/61 % for ADCrel. DWI can facilitate lymph node detection, but alone it is not reliable for differentiating between benign and malignant lymph nodes.