评估应用超声联合弹性成像构建的模型对乙型肝炎显著炎症的诊断性能。

本研究为多中心前瞻性研究。2020年10月至2022年3月从6家医疗机构招募慢性乙型肝炎患者。患者均行肝穿刺活检及超声联合弹性成像检查。依据Scheuer评分评估乙型肝炎炎症活动，将病理炎症程度分为不显著炎症（Scheuer评分≤G1期）和显著炎症（Scheuer评分≥G2期）。将患者分为训练集（来自3个中心）、验证集（来自1个中心）和测试集（来自2个中心）。运用随机森林方法分析生物标志物、超声参数、人口学特征变量，采用逐步回归方法进行特征变量筛选，得出最优子集，纳入Logistc逻辑回归构建模型，绘制模型在3个数据集中诊断乙型肝炎显著炎症（≥G2期）的ROC曲线，并与传统炎症评估方法[丙氨酸氨基转移酶（ALT）、天冬氨酸氨基转移酶（AST）、ALT及AST≥5倍正常值]的性能进行比较。

共792例患者纳入本研究，将其分为训练集463例、验证集166例、测试集163例。总共纳入19个协变量，经筛选确定7个超声特征变量[剪切波传播速度（Vs）、弹性成像（E）、肝脏超声联合弹性成像相关指标A指数（A Index）、门静脉流速、脾脏长度、脾脏厚度、腹壁厚度]纳入Logistic 回归模型，该模型在训练集中的ROC曲线下面积（AUC）为 0.831（95%CI：0.794~0.868），验证集中AUC为0.758（95%CI：0.681~0.834），测试集中AUC为0.726（95%CI：0.645~0.808）。ALT在训练集、验证集、测试集中AUC分别为0.501、0.537、0.540；AST在训练集、验证集、测试集中的AUC分别为0.516、0.436、0.491。ALT及AST≥5倍正常值在训练集、验证集、测试集中AUC分别为0.508、0.519、0.481。

超声联合弹性成像技术在评估乙型肝炎显著炎症中具有较高且稳定的诊断效能，显著优于传统评估方法，可为慢性乙型肝炎患者抗炎治疗提供参考依据。

To evaluate the diagnostic performance of a model constructed based on ultrasound combined with elastography for significant hepatic inflammation in patients with chronic hepatitis B.

This was a multicenter prospective study, in which patients with chronic hepatitis B were recruited from six medical institutions between October 2020 and March 2022. All patients underwent liver needle biopsy and combined ultrasound-elastography examination. Hepatic inflammatory activity was graded according to the Scheuer scoring system, and pathological inflammation was classified as either non-significant (Scheuer score ≤G1) or significant (Scheuer score ≥G2). The patients were divided into a training set (from 3 centers), a validation set (from 1 center), and a test set (from 2 centers). Biomarkers, ultrasound parameters and demographic characteristics were analyzed using the random forest method. Stepwise regression was used for feature selection to obtain the optimal subset, which was incorporated into logistic regression to establish a diagnostic model. Receiver operating characteristic (ROC) curves were generated to evaluate the model's performance in diagnosing significant hepatic inflammation (≥G2) in the three datasets, and its performance was then compared with traditional inflammatory assessment methods, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), and both ALT and AST ≥5 times the upper limit of normal (ULN).

A total of 792 patients were enrolled, including 463 in the training set, 166 in the validation set, and 163 in the test set. Nineteen covariates were initially included, and seven ultrasound features were selected for inclusion in the logistic regression model (named "Best Features"), including shear wave velocity (Vs), elastography (E), A index (liver combined elastography indicator), portal vein flow velocity, splenic length, splenic thickness, and abdominal wall thickness. The area under the ROC curve (AUC) of this model was 0.831 (95% confidence interval [CI]: 0.794–0.868) in the training set, 0.758 (95%CI: 0.681–0.834) in the validation set, and 0.726 (95%CI: 0.645–0.808) in the test set. In the training, validation, and test sets, the AUC values were 0.501, 0.537, and 0.540 for ALT, 0.516, 0.436, and 0.491 for AST, and 0.508, 0.519, and 0.481 for the combination of both ALT and AST ≥5 times ULN.

Ultrasound combined with elastography demonstrates high and stable diagnostic efficacy in evaluating significant hepatic inflammation in chronic hepatitis B, significantly outperforms traditional assessment methods, and can serve as a valuable reference for anti-inflammatory treatment in patients with chronic hepatitis B.