基于CT平扫影像组学模型与系统免疫炎症指数（SII）建立预测急性胰腺炎（AP）复发的模型，并验证其预测效能。

选取阜南县中医院2020年1月至2022年12月收治的145例AP患者，按照7∶3比例分别纳入训练集（n=102）、验证集（n=43）。记录患者随访期间AP复发情况，并比较训练集随访期间发生、未发生AP复发患者CT平扫表现、SII，使用方差阈值法、单变量选择法、最小绝对收缩和选择算子提取并筛选CT平扫影像组学特征，随机森林分类法建立影像组学模型；绘制CT平扫影像组学模型联合SII预测训练集、验证集患者AP复发的受试者工作特征曲线（ROC）并计算曲线下面积（AUC）和预测效能。

训练集102例患者中，29例于随访期间复发，复发率为28.43%；验证集43例患者中，13例复发，复发率为30.23%，复发率差异无统计学意义（P=0.827）。训练集复发、未复发患者AP病因、CT平扫影像组学特征、SII比较，差异有统计学意义（P<0.05）。CT平扫影像组学模型共纳入5项特征，分别为脂肪肝、胸腔积液、肝上间隙积液、肾上腺受侵、胃裸区受侵。以约登指数计算SII最佳截断值为1109.59×10⁹/L。CT影像组学特征联合SII预测训练集、验证集AP复发的AUC分别为0.892、0.837，灵敏度79.31%、76.92%，特异性93.15%、86.67%。

AP患者复发率较高且与病因、CT平扫影像组学特征、SII均存在一定关联，联合CT平扫影像组学特征和SII能够为AP复发的预测提供参考。

To establish a model for predicting the recurrence of acute pancreatitis (AP) based on the model of CT plain scan image and systemic immune inflammatory index (SII), and to verify its predictive efficiency.

A total of 145 patients with AP admitted to Funan County Hospital of Traditional Chinese Medicine from January 2020 to December 2022 were selected and included in the training set (n=102) and the verification set (n=43) according to the ratio of 7∶3. The recurrence of AP during patients' follow-up was recorded, and the CT plain scan performance and SII of patients with and without AP recurrence during the follow-up of the training set were compared. The features of CT plain scan images were extracted and screened using variance threshold method, univariate selection method, minimum absolute contraction and selection operator, and an image omics model was established using random forest classification method. The receiver operating characteristic curve (ROC) was drawn for predicting AP recurrence in patients using CT plain scan imaging omics models combined with SII prediction training and validation sets, and the area under the curve (AUC) and prediction efficiency were calculated.

Among the 102 patients in the training set, 29 patients recurred during the follow-up period, with a recurrence rate of 28.43%. Among the 43 patients in the verification set, 13 cases recurred, with a recurrence rate of 30.23%. There was no statistically significant difference in AP recurrence rate between the training group and the verification group (P=0.827). There were statistically significant differences in AP etiology, CT plain scan imaging omics features and SII between recurrent and non-recurrent patients in the training set (P<0.05). The CT plain scan imaging model included five features, namely fatty liver, pleural effusion, suprahepatic space effusion, adrenal gland invasion and gastric bare area invasion. The best cutoff value of SII calculated by Jordan index was 1 109.59×10⁹/L. The AUC of CT imaging characteristics combined with SII in predicting the recurrence of AP in the training set and the verification set were 0.892 and 0.837, respectively, with sensitivity of 79.31% and 76.92% and specificity of 93.15% and 86.67%, respectively.

The recurrence rate of AP patients is high, and it is related to the etiology, CT imaging features and SII. The combination of CT imaging features and SII can provide reliable reference for the prediction of AP recurrence.