目的 探讨免疫炎症指标及磁控胶囊内镜特点在胃腺癌及癌前病变鉴别诊断中的价值，构建并验证风险预测模型。 方法 回顾性分析2021年1月至2023年12月在河北省人民医院接受磁控胶囊内镜检查的578例患者的病历资料。依据80/20定律随机分为训练集（462例）和验证集（116例）。行磁控胶囊内镜检查及血细胞检测，以病理诊断结果为“金标准”，分为胃腺癌组与癌前病变组。比较胃腺癌与癌前病变患者的磁控胶囊内镜特点、中性粒细胞与淋巴细胞比（neutrophil-lymphocyte ratio， NLR）、血小板与淋巴细胞比（platelet-lymphocyte ratio， PLR），构建并验证胃腺癌的风险诊断模型。 结果 462例磁控胶囊内镜检查患者中，经病理检查确诊胃腺癌76例，占比16.45%（76/462）；癌前病变386例，占比83.55%（386/462）。验证集116例胃内镜检查患者中，胃腺癌22例，占比18.97%（22/116）；癌前病变94例，占比81.03%（94/116）。两组患者的病变大小、病灶边界、黏液、病变形态比较，差异无统计学意义（P > 0.05）；胃腺癌组患者中颜色发白占比、表面微结构不规则占比、不规则网格状微血管占比高于癌前病变组（P < 0.05）。胃腺癌组患者的NLR、PLR高于癌前病变组（P < 0.05）。表面微结构不规则（OR = 2.213，95%CI：1.288 ~ 3.801）、不规则网格状微血管（OR = 2.489，95%CI：1.458 ~ 4.249）、NLR（OR = 2.369，95%CI：1.389 ~ 4.046）、PLR（OR = 3.016，95%CI：1.767 ~ 5.148）是胃腺癌的危险因素（P < 0.05）。风险模型诊断训练集胃腺癌的灵敏度为0.800（95%CI：0.716 ~ 0.891），特异度为0.783（95%CI：0.694 ~ 0.851），曲线下面积为0.858（95%CI：0.787 ~ 0.931）。风险模型诊断验证集胃腺癌的灵敏度为0.861（95%CI：0.771 ~ 0.945），特异度为0.769（95%CI：0.683 ~ 0.841），曲线下面积为0.844（95%CI：0.765 ~ 0.923）。 结论 胃损伤病灶的表面微结构、微血管形态及NLR、PLR与胃腺癌的发病有关，构建风险诊断模型有助于早期鉴别诊断胃腺癌。

Objective To investigate the potential of immune?inflammatory markers and the characteristics of magnetically controlled capsule endoscopy in distinguishing gastric adenocarcinoma from precancerous lesions， as well as to develop and validate a risk prediction model. Methods Retrospective analysis was conducted on medical records of 578 patients who underwent magnetic controlled capsule endoscopy at our hospital between January 2021 and December 2023. Following the principle of Pareto's law （80/20 rule）， they were randomly divided into a training set （462 cases） and a validation set （116 cases）. Magnetic controlled capsule endoscopy and blood cell tests were performed， with pathological diagnosis results serving as the "gold standard"， to classify patients into groups of gastric adenocarcinoma and precancerous lesions. The magnetic controlled capsule endoscopic features， neutrophil?lymphocyte ratio （NLR）， platelet?lymphocyte ratio （PLR） in patients with gastric adenocarcinoma and precancerous lesions were compared to develop and validate a risk diagnostic model for gastric adenocarcinoma. Results Among the 462 patients who underwent magnetic controlled capsule endoscopy， gastric adenocarcinoma was diagnosed in 76 cases through pathological examination， accounting for 16.45% （76/462）， while precancerous lesions were observed in 386 cases， accounting for 83.55% （386/462）. In the validation set of 116 patients who underwent gastric endoscopy， there were 22 cases of gastric adenocarcinoma， representing an incidence rate of 18.97% （22/116）， and a total of 94 cases with precancerous lesions， accounting for an incidence rate of 81.03% （94/116）. No statistically significant differences （P > 0.05） were found between the two groups regarding lesion size， border appearance， mucus presence or lesion morphology. However， compared to the precancerous lesion group， the proportion of whitish coloration as well as irregular surface microstructure and grid?like microvessels was significantly higher in the gastric adenocarcinoma group （P < 0.05）. Moreover， both NLR and PLR values were significantly higher in the gastric adenocarcinoma group compared to those in the precancerous lesion group （P < 0.05）. Irregular surface microstructure （OR = 2.213， 95%CI： 1.288 ~ 3.801）， irregular grid?like microvessels （OR = 2.489， 95%CI： 1.458 ~ 4.249）， NLR （OR = 2.369， 95%CI： 1.389 ~ 4.046）， and PLR （OR = 3.016， 95%CI： 1.767 ~ 5.148） were identified as risk factors for gastric adenocarcinoma （P < 0.05）. The sensitivity of the risk model for diagnosing gastric adenocarcinoma in the training set was 0.800 （95% CI： 0.716 ~ 0.891）， with a specificity of 0.783 （95% CI： 0.694~0.851） and an area under the curve of 0.858 （95% CI： 0.787 ~ 0.931）. In the validation set， the sensitivity for diagnosing gastric adenocarcinoma was 0.861 （95% CI： 0.771 ~ 0.945）， with a specificity of 0.769 （95% CI： 0.683~0.841） and an area under the curve of 0.844 （95% CI： 0.765 ~ 0.923）. Conclusion The surface microstructure， microvascular morphology， NLR， and PLR of gastric lesions are correlated with the occurrence of gastric adenocarcinoma. Developing a risk diagnostic model facilitates early identification and diagnosis of gastric adenocarcinoma.