Published in:
01-01-2019
Skewness of apparent diffusion coefficient (ADC) histogram helps predict the invasive potential of intraductal papillary neoplasms of the bile ducts (IPNBs)
Published in: Abdominal Radiology | Issue 1/2019
Login to get accessAbstract
Objective
This retrospective study was to explore the value of whole lesion apparent diffusion coefficient (ADC) histogram in distinguishing invasive and noninvasive intraductal papillary neoplasms of the bile ducts (IPNBs).
Method and materials
Fifty-two patients of IPNB underwent MRI at 1.5T with diffusion-weighted imaging (DWI, b = 500 s/mm2) before surgical resections. ADC histogram metrics were generated by using the software MR OncoTreat. The mean, standard deviation, median, skewness, kurtosis as well as the 10th, 25th, 75th, and 90th percentiles were compared between pathologically defined invasive (n = 35) and noninvasive (n = 17) IPNBs. Such conventional imaging characters as lesion location, bile duct wall dilation, and mural nodularity were also assessed. Multivariate regression analysis as well as receiver operating characteristics (ROC) analysis were then conducted to determine the predictive factors and to evaluate potential diagnostic performances.
Results
The inter-operator reliability was good to excellent (ICC: 0.693–979). Mean median, kurtosis, and the 10th, 25th, 75th, 90th percentiles were all greater in noninvasive group than invasive ones (P: 0.00–002). Skewness was lower in noninvasive group than invasive ones (− 1.0 ± 0.6 vs. − 0.3 ± 0.6, P = 0.00). After multivariate regression, skewness (AUC = 0.822, 95%CI 0.70–0.91) and mural nodularity (accuracy = 0.808) were the only two independent factors in predicting invasive IPNBs. The diagnostic performance improved (AUC = 0.867, 95%CI 0.742–0.946) when combining skewness and mural nodularity, however, the difference did not reach statistical significance (P = 0.16).
Conclusion
The ADC histogram has capability of distinguishing invasive and noninvasive IPNBs, in which skewness was an independent predictive factor.