Preoperative identification of cytokeratin 19 status of hepatocellular carcinoma based on diffusion kurtosis imaging

Purpose To explore the potential value of diffusion kurtosis imaging (DKI) for identification of cytokeratin 19 (CK19) status of HCCs. Methods This study was approved by the local institute review board and written informed consent was obtained. 73 patients with pathologically confirmed HCCs were included in this prospective study. All the diffusion-weighted (DW) images were acquired using a 3.0-T MR scanner with 4 b-values (0, 800, 1500 and 2000 s/mm(2)). The mean diffusion value (MD) and mean kurtosis coefficient (MK) from DKI, apparent diffusion coefficient (ADC) from DW imaging (b = 0, 500 s/mm(2)), and tumor-to-liver signal intensity ratios on ADC map (SIRADC) and DW images with b-value of 500 s/mm(2) (SIRb500) were calculated and compared between CK19-positive (n = 23) and CK19-negative (n = 50) HCC groups. Univariate and multivariate logistic regression analyses were used to identify risk factors for the positive expression of CK19. Results Increased a-fetoprotein level (p = 0.021) and SIRb500 (p = 0.006) and decreased ADC (p = 0.021) and MD (p < 0.001) were significantly correlated with CK19-positive HCCs at univariate analysis. Decreased MD value (odds ratio: 0.042, p = 0.002) and a-fetoprotein level (odds ratio: 5.139, p = 0.015) were the independent risk factors for CK19-positive HCCs at multivariate analysis. The area under the curve of MD value by receiver operating characteristic analysis was 0.823 with a sensitivity of 86.96% and a specificity of 76% for the prediction of CK19-positive HCCs. Conclusion The decreased MD value derived from DKI is potential quantitative biomarker for predicting CK19-positive HCCs.