Alempijevic Tamara, Zec Simon, Nikolic Vladimir, Veljkovic Aleksandar, Stojanovic Zoran, Matovic Vera, Milosavljevic Tomica
School of Medicine, University of Belgrade, Clinic for Gastroenterology and Hepatology, Clinical Center of Serbia, Serbia.
School of Medicine, University of Belgrade, Serbia.
Med Ultrason. 2017 Jan 31;19(1):7-15. doi: 10.11152/mu-921.
Accurate clinical assessment of liver fibrosis is essential and the aim of our study was to compare and combine hemodynamic Doppler ultrasonography, liver stiffness by transient elastography, and non-invasive serum biomarkers with the degree of fibrosis confirmed by liver biopsy, and thereby to determine the value of combining non-invasive method in the prediction significant liver fibrosis.
We included 102 patients with chronic liver disease of various etiology. Each patient was evaluated using Doppler ultrasonography measurements of the velocity and flow pattern at portal trunk, hepatic and splenic artery, serum fibrosis biomarkers, and transient elastography. These parameters were then input into a multilayer perceptron artificial neural network with two hidden layers, and used to create models for predicting significant fibrosis.
According to METAVIR score, clinically significant fibrosis (≥F2) was detected in 57.8% of patients. A model based only on Doppler parameters (hepatic artery diameter, hepatic artery systolic and diastolic velocity, splenic artery systolic velocity and splenic artery Resistance Index), predicted significant liver fibrosis with a sensitivity and specificity of75.0% and 60.0%. The addition of unrelated non-invasive tests improved the diagnostic accuracy of Doppler examination. The best model for prediction of significant fibrosis was obtained by combining Doppler parameters, non-invasive markers (APRI, ASPRI, and FIB-4) and transient elastography, with a sensitivity and specificity of 88.9% and 100%.
Doppler parameters alone predict the presence of ≥F2 fibrosis with fair accuracy. Better prediction rates are achieved by combining Doppler variables with non-invasive markers and liver stiffness by transient elastography.
对肝纤维化进行准确的临床评估至关重要。本研究旨在将血流动力学多普勒超声检查、瞬时弹性成像测定的肝脏硬度以及非侵入性血清生物标志物与经肝活检证实的纤维化程度进行比较和结合，从而确定联合使用非侵入性方法预测显著肝纤维化的价值。
我们纳入了102例病因各异的慢性肝病患者。对每位患者进行门静脉主干、肝动脉和脾动脉的速度及血流模式的多普勒超声测量、血清纤维化生物标志物检测以及瞬时弹性成像检查。然后将这些参数输入具有两个隐藏层的多层感知器人工神经网络，用于创建预测显著纤维化的模型。
根据METAVIR评分，57.8%的患者检测到临床显著纤维化（≥F2）。仅基于多普勒参数（肝动脉直径、肝动脉收缩期和舒张期速度、脾动脉收缩期速度和脾动脉阻力指数）建立的模型预测显著肝纤维化的敏感性和特异性分别为75.0%和60.0%。添加不相关的非侵入性检查提高了多普勒检查的诊断准确性。通过结合多普勒参数、非侵入性标志物（APRI、ASPRI和FIB-4）以及瞬时弹性成像获得了预测显著纤维化的最佳模型，其敏感性和特异性分别为88.9%和100%。
仅多普勒参数就能以合理的准确性预测≥F2纤维化的存在。将多普勒变量与非侵入性标志物以及瞬时弹性成像测定的肝脏硬度相结合可获得更好的预测率。
