Apparent diffusion coefficient (ADC) in discriminating between solid benign and malignant focal liver lesions (FLLs) with diffusion-weighted imaging (DWI): comparison of ADC thresholding versus normalized ADC thresholding and lesion-to-liver ADC ratio

Purpose: The purpose of this study was to compare the diagnostic performance of three different methods of ADC estimation in differentiating solid benign and malignant FLLs. Material and Methods: We included forty-five patients with 75 FLLs detected on a 1.5 T system, and proven to be malignant and benign in 55 and 20 cases, respectively (cysts/haemangiomas excluded). During separate reading sessions, two readers in consensus evaluated DWI images and the ADC map. FLLs were assessed as benign or malignant: (a) according to the threshold established by a receiver operating characteristic (ROC) analysis, without (ADC-T) and with (ADC-TS) normalization for the spleen, respectively; (b) by calculating the lesion-to-liver ADC ratio (ADC-R; cut-off <1). End-points for the three methods were the positive- and negative-predictive value (PPV, NPV) for malignancy and the area under the curve (AUC) values. Results: ADC threshold for malignancy was 1.06 x10-3 mm2/sec for ADC-T and 1.27 x10-3 mm2/sec for ADC-TS. ADC-T, ADC-TS and ADC-R showed high PPV (89.2%, 84.6% and 78.1%, respectively), but low NPV (42.1%, 38.9% and 30.2%, respectively). AUCs of ADC-T (0.70) and ADC-TS (0.65) were higher as compared to that of ADC-R (0.55), showing a significant difference between ADC-T and ADC-R (p<0.01). Main source of false-negative cases was hepatocarcinoma (22-30/55). Conclusion: Normalizing FLLs ADC for the spleen or providing the ADC-R did not improve the diagnostic performance of ADC-T, which is limited by a low NPV.