Open AccessValue of MR histogram analyses for prediction of microvascular invasion of hepatocellular carcinoma
Author(s) -
Yunpeng Huang,
He-Yue Liang,
Zhaoxia Yang,
Ying Ding,
Mengsu Zeng,
Sheng-Xiang Rao
Publication year - 2016
Publication title -
medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.59
H-Index - 148
eISSN - 1536-5964
pISSN - 0025-7974
DOI - 10.1097/md.0000000000004034
Subject(s) - medicine , hepatocellular carcinoma , value (mathematics) , histogram , radiology , oncology , artificial intelligence , statistics , image (mathematics) , mathematics , computer science
The objective is to explore the value of preoperative magnetic resonance (MR) histogram analyses in predicting microvascular invasion (MVI) of hepatocellular carcinoma (HCC). Fifty-one patients with histologically confirmed HCC who underwent diffusion-weighted and contrast-enhanced MR imaging were included. Histogram analyses were performed and mean, variance, skewness, kurtosis, 1th, 10th, 50th, 90th, and 99th percentiles were derived. Quantitative histogram parameters were compared between HCCs with and without MVI. Receiver operating characteristics (ROC) analyses were generated to compare the diagnostic performance of tumor size, histogram analyses of apparent diffusion coefficient (ADC) maps, and MR enhancement. The mean, 1th, 10th, and 50th percentiles of ADC maps, and the mean, variance. 1th, 10th, 50th, 90th, and 99th percentiles of the portal venous phase (PVP) images were significantly different between the groups with and without MVI ( P <0.05), with area under the ROC curves (AUCs) of 0.66 to 0.74 for ADC and 0.76 to 0.88 for PVP. The largest AUC of PVP (1th percentile) showed significantly higher accuracy compared with that of arterial phase (AP) or tumor size ( P <0.001). MR histogram analyses—in particular for 1th percentile for PVP images—held promise for prediction of MVI of HCC.