Contrast-enhanced ultrasound combined with mathematical pattern identify intrahepatic cholangiocarcinoma and prophase of bacterial hepatic abscess / 中华超声影像学杂志

ABSTRACT

Objective:To explore the application of contrast-enhanced ultrasound (CEUS) combined with mathematical model in differential diagnosis of intrahepatic cholangiocarcinoma (ICC) and prophase of bacterial hepatic abscess (PBHA).Methods:Fifty-one ICC patients (ICC group)and 46 PBHA patients(PBHA group) from January 2010 to April 2019 in Henan Provincial People′s Hospital were analyzed retrospectively. All ICC patients and 41 PBHA patients were confirmed by surgery or puncture pathology, and 5 PBHA patients were confirmed by clinical follow-up. The characteristic information of ultrasound images of the lesions were analyzed.Partial least squares-discriminant analysis (PLS-DA) was used to establish ICC and PBHA mathematical models, and Monte Carlo simulation was used to verify their accuracy. Based on PLS-DA modeling method, variable importance in the projection (VIP) was used to obtain the variables that had a strong influence on identifying the two from all variables.Results:There were significant differences between ICC group and PBHA group in lesion sites, complications, internal characteristics and clearance characteristics of contrast-enhanced arterial phase (all P<0.05). The average positive and negative predictive values of PLS-DA were 92.30% and 91.17%, and the average positive and negative predictive values in the prediction set were 100% and 94.11%, respectively.Variables with VIP value >1 included X2 (complication), X4, X5, X7, X8, X9 (CEUS enhanced features), X10, X11 (clearing features), and these variables could be used as important indicators for differential diagnosis between ICC and PBHA. Conclusions:PLS-DA method based on CEUS parameters can construct a differential diagnosis model for ICC and PHBA, which is expected to provide a valuable and robust diagnostic method for these two diseases which are easily confused and lack of specific imaging manifestations.