Accuracy of visual assessment versus apparent diffusion coefficient quantification in differentiating malignant and solid benign focal liver lesions with diffusion-weighted imaging

Purpose: The purpose of this study was to compare the accuracy of diffusionweighted imaging (DWI) visual analysis (VA) versus the apparent diffusion coefficient quantification (ADC-Q) in assessing malignancy of solid focal liver lesions (FLLs). Material and Methods: Two radiologists in consensus retrospectively assessed as benign or malignant 50 FLLs (16 benign, 24 malignant) on 32 patients examined on a 1.5 T. Two different methods were used in separate reading sessions: (a) VA of signal intensity on DWI images at b=800 sec/mm2 and ADC-map; (b) lesion ADC measurement on the ADC-map. Reference standard included histology and follow-up confirmation of a consensus panel diagnosis. We estimated the accuracy for malignancy of both methods and the ADC-Q threshold as assessed by a receiver operating characteristic (ROC) analysis. Results: Because of 20 false-negative hepatocarcinoma, VA showed lower accuracy than ADC-Q (52.0% vs. 68.0%). However, stratified accuracy for metastases was higher than VA (75.0 vs. 66.7%). ADCs and signal features of malignant and benign FLLs largely overlapped. Conclusion: VA performed worse than ADC-Q for hepatocarcinoma, and better for metastases, possibly in relation with the T2-shine-through phenomenon. Overall, the accuracy of both methods was limited because of the overlap in visual appearance and ADC values between solid benign and malignant FLLs.