原发性胃癌淋巴结转移风险的新型目标预测模型

A novel target prediction model for lymph node metastasis of primary gastric cancer

目的寻找原发性胃癌淋巴结转移的危险因素,并构建新型目标预测模型。方法采用回顾性队列研究的方法,收集2019年12月—2022年6月在本院普外科和肿瘤外科行胃癌根治术的514例原发性胃癌患者的临床病理资料,所有患者被随机分为目标模型组(347例,67.5%)和对比验证组(167例,32.5%)。先运用Logistic单因素分析方法,对所有病例样本进行回顾性分析,筛查影响淋巴结转移的变量,确定目标预测模型输入节点的变量项目,再使用多层感知器训练目标模型。目标模型由Logistic单因素分析筛选出的变量构成输入层。人工智能依据输入数据分析患者淋巴结转移状态,并与真实值进行比较。通过绘制受试者操作特性(ROC)曲线、获取曲线下面积(AUC)来评估模型的准确性。结果目标模型组与对比验证组临床资料的比较,差异均无统计学意义(均P>0.05)。多因素分析结果显示,中性粒细胞与淋巴细胞比值(NLR)、肿瘤位置、肿瘤形态、肿瘤长径、分化程度、以及肿瘤分期与淋巴结转移有关。根据目标预测模型绘制出的ROC曲线分析显示,目标模型组的AUC为0.841(95%CI:0.795-0.863),对比验证组的AUC为0.707(95%CI:0.679-0.852),提示目标预测模型有较高的淋巴结转移诊断能力。结论原发性胃癌淋巴结转移与中性粒细胞与淋巴细胞比值、肿瘤位置、肿瘤形态、肿瘤长径、肿瘤分化程度、肿瘤分期等有关。本研究构建的新型目标预测模型能准确预测原发性胃癌患者的淋巴结转移情况,具有临床诊断和治疗及预后评估价值。

Objective To find out the risk factors of lymph node metastasis of primary gastric cancer and construct a new target prediction model.Methods Retrospective cohort study was adopted.The clinical data of 514 patients with primary gastric cancer who underwent radical gastrectomy in the general surgery department and tumor surgery department of the first affiliated hospital of Bengbu Medical College from December 2019 to June 2022 were collected.All patients were randomly divided into the target model group(347 patients,67.5%)and the comparison validation group(167 patients,32.5%).Single factor Logistic analysis method was used to conduct retrospective analysis on all case samples to screen variables affecting lymph node metastasis,then determine the variable items of the target prediction model input node,and then use multi-layer perceptron to train the target model.The target model was composed of the input layer of variables screened by single factor logistic analysis.AI analyzes patients'lymph node metastasis status based on input data and compares it with the real value.The accuracy of the model was evaluated by drawing the ROC curve and obtaining the area under the curve(AUC).Results There was no significant difference in the clinical data between the target model group and the contrast validation group(P>0.05).Multivariate analysis showed that the neutrophils to lymphocytes ratio(NLR),tumor location,tumor shape,tumor length,differentiation,and tumor stage were related to lymph node metastasis.According to the ROC curve analysis drawn by the target prediction model,the AUC of the target model group was 0.841(95%CI:0.795-0.863),and the AUC of the comparison verification group was 0.707(95%CI:0.679-0.852),suggesting that the target prediction model has a higher ability to diagnose lymph node metastasis.Conclusions Lymph node metastasis of primary gastric cancer is related to the neutrophils to lymphocytes ratio,tumor location,tumor shape,tumor length,tumor differentiation and tumor stage.The novel target prediction model constructed in this study can accurately predict lymph node metastasis in patients with primary gastric cancer,and has clinical diagnosis,treatment and prognosis evaluation value.