Comparison of three methods of apparent diffusion coefficient evaluation in assessing solid focal liver lesions with diffusion-weighted imaging (DWI)

Purpose: The purpose of this study was to compare the performance in diagnosing focal liver lesions (FLLs) malignancy of the apparent diffusion coefficient thresholding (ADC-T) method versus two variants of the lesion-toliver ADC ratio (ADC-R). Material and Methods: Examinations were performed on a 1.5 T system. Analysis was applied to 50 FLLs proven to be malignant and benign in 34 and 16 cases, respectively. Cysts and haemangiomas were excluded from the analysis. We estimated the positive-predictive value (PPV) and negativepredictive value (NPV) for malignancy of ADC-T and two variants of the ADC-R, calculated without (ADC-R1) and with (ADC-R2) the inclusion of the parenchymal ADC standard deviation value, respectively. An ADC-R < 1 was considered as malignant. A receiver operating characteristic (ROC) analysis was performed to establish ADC-T threshold for malignancy and to compare the areas under the curve (AUC). Results: The ADC-T threshold for malignancy was ≤1.09x10-3 mm2/sec. PPV and NPV were 87.5% (95%C.I.: 74.4-94.7) and 50.0% (95%C.I.: 35.7-64.3) for ADC-T versus 75.0% (95%C.I.: 60.4-85.7) and 36.7% (95%C.I.: 23.9-51.5) for ADC-R1 versus 90.9% (95%C.I.: 78.5-96.8) and 38.5% (95%C.I.: 25.4-53.3) for ADC-R2, respectively. Most false-negative FLLs were represented by hepatocellular carcinoma. ADC-T showed an AUC (0.71) significantly higher (p<0.05) than ADC-R1 (0.56), but not than ADC-R2 (0.61).Conclusion: Regardless of the ADC evaluation method, NPV was low, while the PPV was high, especially by comparing lesion and parenchymal ADCs with ADC-R2. However, the use of an ADC threshold provided higher diagnostic performance in terms of AUC.