基于人工智能系统的CT定量分析诊断肺结节浸润性的价值

Value of CT quantitative analysis based on artificial intelligence system in diagnosing infiltration of pulmonary nodules

目的探讨人工智能(AI)量化参数评估肺结节浸润程度的临床价值。方法回顾性分析2021年11月至2023年9月在亳州市人民医院住院治疗的114例经手术病理证实的肺腺癌患者临床资料,并根据病理结果分为非浸润组(n=72)、浸润组(n=42)。比较两组一般资料和AI量化参数之间的差异,采用logistic回归分析影响肺腺癌浸润程度的因素,以受试者工作特征(ROC)曲线评估量化参数对肺腺癌浸润程度的预测价值。结果两组一般资料(年龄、性别)和12个量化参数(熵、CT平均值、CT最小值、3D长径、体积、质量、长短径平均值、最大面面积、偏度、CT最大值、CT值方差和表面积)比较,差异均有统计学意义(P<0.05)。logistic回归结果显示,CT平均值、熵是影响肺腺癌发生浸润性的独立危险因素,CT最小值是影响肺腺癌发生浸润性的独立保护因素(P<0.05)。ROC结果显示,CT平均值、熵和CT最小值预测肺腺癌发生浸润性的曲线下面积(AUC)分别为0.630、0.888、0.890,3者联合预测的AUC为0.955,优于各自单独检测(P<0.05)。结论CT量化参数CT平均值、熵及CT最小值水平可为临床诊断肺腺癌浸润性提供帮助,3项指标联合检测可进一步提高诊断价值。

Objective To investigate the clinical value of quantitative artificial intelligence(AI)parameters to assess the degree of pul⁃monary nodal infiltration.Methods The clinical data of 114 patients with lung adenocarcinoma confirmed by surgical pathology who were hospitalised in Bozhou People's Hospital from November 2021 to September 2023 were retrospectively analysed and divided into the noninfiltrating group(n=72)and infiltrating group(n=42)according to the pathology results.Differences in general information and quantitative pa⁃rameters of AI were compared between the two groups,and logistic regression was used to analyse the factors affecting the degree of lung adeno⁃carcinoma infiltration,and the predictive value of quantitative parameters on the degree of lung adenocarcinoma infiltration was assessed by the receiver operator characteristic(ROC)curve.Results The differences in general information(age,gender)and 12 quantitative parameters(en⁃tropy,mean CT value,minimum CT value,3D longitudinal diameter,volume,mass,mean long and short diameters,maximum surface area,skewness,maximum CT value,variance of CT value and surface area)between the two groups were statistically significant(P<0.05).The logistic regression results showed that the mean CT value and entropy were independent risk factorsfor the development of infiltrative lung adenocarci⁃noma,and the minimum CT value was an independent protective factor for the development of infiltrative lung adenocarcinoma(P<0.05).The ROC results showed that the area under the curve(AUC)of the three predicted the occurrence of infiltrative lung adenocarcinoma was 0.630,0.888,and 0.890,respectively,and the AUC predicted by the combination of the three was 0.955,which was superior to that of their individual detection(P<0.05).Conclusions The quantitative CT parameters meanCT value,entropy and minimum CT levels can provide clinical diagno⁃sis of lung adenocarcinoma infiltration,and the combined detection of the three indexes can further improve the diagnostic value.