【摘要】 目的：基于病理结果探讨人工智能（AI）联合CT影像特征评估肺腺癌侵袭性的价值。方法：选取应城市人民医院2020年12月至2022年12月诊断为肺腺癌的108例肺结节患者为研究对象,根据病理结果分为浸润前/微浸润组（30例）和浸润组（78例）,比较两组之间临床特点、CT影像特征和AI系统参数的差异,利用Logistic回归筛选浸润性肺腺癌的独立预测因素,分别构建CT影像模型、AI模型及联合模型,采用受试者工作特征曲线（ROC）评估模型诊断效能,并利用DeLong检验进行比较。结果：浸润组患者年龄更大、男性比例更高,以实性结节为主,具有分叶、毛刺及胸膜凹陷征等特征,结节的长、短径较大,AI风险分级为高危的比例更高（P <0.05）。Logistic回归分析显示,年龄、结节类型、毛刺征是浸润性肺腺癌的独立预测因素（P <0.05）。CT影像模型、AI模型及联合模型诊断浸润性肺腺癌的ROC曲线下面积分别为0.892、0.823、0.915,联合模型诊断效能显著优于单独的CT影像模型和AI模型（P <0.05）。结论：AI能够有效提升传统CT影像诊断浸润性肺腺癌的效能,可作为肺腺癌侵袭性评估的重要辅助工具。更多还原

【Abstract】 Objective To explore the value of Artificial Intelligence(AI) integrated with CT image features in evaluating the invasiveness of lung adenocarcinoma based on pathological results. Methods A total of 108 patients with pulmonary nodules diagnosed as lung adenocarcinoma in Yingcheng City People’s Hospital from December 2020 to December 2022 were selected as the study subjects, which were divided into pre-infiltrating/micro-infiltrating group(30 cases) and infiltrating group(78 cases) based on the pathological results. Differences in clinical features, CT imaging features and AI system parameters between the two groups were compared, and logistic regression was used to screen the independent predictors of invasive lung adenocarcinoma. CT imaging model, AI model and integrated model were constructed respectively. Receiver Operating Characteristic curve(ROC) was used to evaluate the diagnostic effectiveness of the models, and DeLong test was used for comparison. Results In the infiltrating group, the patients were older, with a higher proportion of males, mainly solid nodules with lobed, burr and pleural depression signs, and larger long and short diameters of nodules, and higher proportion of AI risk classified as high-risk(P<0.05). Logistic regression analysis showed that age, nodule type, and burr sign were independent predictors of infiltrating lung adenocarcinoma(P<0.05). The areas under the ROC curves of the CT imaging model, AI model, and integrated model in diagnosing invasive lung adenocarcinoma were 0.892、0.823 and 0.915 respectively, and the diagnostic effectiveness of the integrated model was significantly better than that of single CT imaging model or AI model(P<0.05). Conclusion AI can effectively improve the effectiveness of traditional CT imaging in the diagnosis of invasive lung adenocarcinoma,and can be used as an important auxiliary tool in the evaluation of invasive lung adenocarcinoma.更多还原