Can diffusion-weighted imaging distinguish between benign and malignant pediatric liver tumors?

There are limited data on utility of diffusion-weighted imaging (DWI) in the evaluation of pediatric liver lesions.

To determine whether qualitative and quantitative DWI can be used to differentiate benign and malignant pediatric liver lesions.

We retrospectively reviewed MRIs in children with focal liver lesions to qualitatively evaluate lesions noting diffusion restriction, T2 shine-through, increased diffusion, hypointensity on DWI and apparent diffusion coefficient (ADC) maps, and intermediate signal on both, and to measure ADC values. Pathology confirmation or a combination of clinical, laboratory and imaging features, and follow-up was used to determine final diagnosis.

We included 112 focal hepatic lesions in 89 children (median age 11.5 years, 51 female), of which 92 lesions were benign and 20 malignant. Interobserver agreement was almost perfect for both qualitative (kappa 0.8735) and quantitative (intraclass correlation coefficient [ICC] 0.96) diffusion assessment. All malignant lesions showed diffusion restriction. Most benign lesions other than abscesses were not restricted. There was significant association of qualitative restriction with malignancy and non-restriction with benignancy (Fisher exact test P<0.0001). Mean normalized ADC values of malignant lesions (1.23x10^-3 mm²/s) were lower than benign lesions (1.62x10^-3 mm²/s; Student's t-test, P<0.015). However, there was significant overlap of ADC between benign and malignant lesions, with wide range for each diagnosis. Receiver operating characteristic (ROC) analysis revealed an area under the curve (AUC) of 0.63 for predicting malignancy using an ADC cut-off value of ≤1.20x10^-3 mm²/s, yielding a sensitivity of 78% and a specificity of 54% for differentiating malignant from benign lesions.

Qualitative diffusion restriction in pediatric liver lesions is a good predictor of malignancy and can help to differentiate between benign and malignant lesions, in conjunction with conventional MR sequences. Even though malignant lesions demonstrated significantly lower ADC values than benign lesions, the use of quantitative diffusion remains limited in its utility for distinguishing them because of the significant overlap and wide ranges of ADC values.