Evaluation Of Venous Invasion In Hepatic Alveolar Echinococcosis Based On MDCT And Radiomics

Objective:Alveolar echinococcosis(AE)is a rare and fatal zoonotic parasitic disease that exhibits insidious growth with tumor-like invasive characteristics and potential for metastasis.It easily invades important adjacent blood vessels and bile ducts.Vascular invasion status directly affects the selection of surgical methods,so it is very important to accurately evaluate the status of hepatic AE vascular invasion for clinical treatment decision.At present,the clinical studies on the evaluation of arterial abnormalities by computed tomography angiography(CTA)are relatively mature,but the imaging criteria for the evaluation of venous invasion is still a hot topic.Because venous imaging is different from hepatic artery imaging,the image quality is often affected by many factors,and clear venous imaging is the basis for evaluating venous invasion.In order to improve the image quality of portal vein,contrast enhancement boost(CE-boost)is used in multi-detector computed tomography(MDCT).CE-boost technology can improve the image quality of portal vein,in order to improve the visualization of portal vein in enhanced abdominal scan.In addition,clinical data from surgical cases suggest that existing MDCT criteria or signs for diagnosing tumor vascular invasion may not apply well to hepatic AE due to its unique characteristics regarding vascular invasion.Therefore,this study aims to compare classic MDCT criteria and signs of vascular invasion with the intraoperative and postoperative histopathological results in hepatic AE patients,in order to determine reliable MDCT standards for diagnosing venous invasion in hepatic AE.Furthermore,using non-invasive MDCT examination based on portal venous phase images as a basis for radiomics analysis combined with clinical-routine imaging features will establish a model to assess the clinical value of predicting venous invasiont in hepatic AE.This will lay the foundation for achieving convenient homogeneous evaluation of vascular invasion that can be easily implemented clinically.Methods:(1)A total of 100 hospitalized patients who underwent abdominal MDCT examination according to clinical needs from June 2021 to September 2021 were retrospectively collected.50 patients in group A underwent conventional abdominal CT enhanced examination,and 50 patients in group B underwent abdominal CTA examination.Subjective and objective evaluation of the three groups of images(group A CE-boost,group A without CE-boost and group B)portal vein imaging of different branches were statistically analyzed.(2)From January 2011 to December 2018,136 hospitalized patients with hepatic AE confirmed by histopathology and underwent MDCT examination were retrospectively collected to determine the cut-off value of the lesion-vessel contact angle for the diagnosis of venous invasion.The diagnostic efficacy of lesion-vessel contact angle,vascular lumen stenosis,vascular wall irregularity and vascular lumen occlusion were evaluated.(3)The portal venous phase images of abdominal enhanced CT scans of 259 patients with hepatic AE from January 2012 to December 2021 were retrospectively collected to establish a radiomics model.Combined with clinical and conventional imaging features,a model for combined diagnosis of hepatic AE venous invasion was established.The Receiver operating characteristic(ROC)curve was used to evaluate the efficacy and clinical application value of each model.Results:(1)The signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)of the liver parenchyma and most of the target veins in group A were the highest.The CNR of most portal veins without CE-boost in group A was the lowest(P < 0.001).There was no significant difference in SNR of portal vein between group A without CE-boost and group B(P > 0.05).(2)The lesion-vessel contact Angle > 180° was the cut-off value for the diagnosis of hepatic AE vein invasion,and its diagnostic efficiency was the best.The area under the ROC curve(AUC)was 0.907,95% confidence interval(CI)was 0.872-0.941.The diagnostic efficiency of vascular wall irregularity and vascular lumen stenosis was the lowest.When all the criteria and signs were combined,the diagnostic efficacy was the highest(AUC=0.932,95%CI,0.905-0.959,P < 0.001).(3)Three Machine learning classifiers for radiomics include Logistic regression(LR),Random Forest(RF)and Support Vector Machine(SVM).In the combined model,LR classifier had the best diagnostic performance in the diagnosis of hepatic AE venous invasion,with AUC of 0.9193 and 0.9375 in the training and test datasets,respectively.Conclusion:(1)CE-boost technology can significantly improve the image quality of portal vein imaging without additional scanning settings and contrast agent dose.It does not need to change the previous clinical workflow.(2)The lesion-vessel contact Angle > 180° has the best sensitivity and specificity in the diagnosis of hepatic AE vein invasion,and the interobserver agreement is good.The diagnostic performance of lesion-vessel contact angle > 180° is further improved with the presence of vascular lumen occlusion,vascular wall irregularity or vascular lumen stenosis.(3)The combined model based on radiomics model and clinical-conventional imaging model is the best prediction model for hepatic AE venous invasion before operation,which can provide important reference information for the formulation of clinical diagnosis and treatment plan in a non-invasive way.