人工智能辅助诊断系统预测肺腺癌表皮生长因子受体突变的应用价值

Application Value of Artificial Intelligence-Assisted Diagnosis System in Predicting Epidermal Growth Factor Receptor Mutations in Lung Adenocarcinoma

目的探讨人工智能辅助诊断系统基于CT图像预测肺腺癌表皮生长因子受体突变的价值,为临床诊断或决策提供支持。方法回顾性分析2017至2019年就诊于我院经手术或穿刺活检病理证实为肺腺癌患者143例,根据基因检测结果将其分为EGFR突变组(n=68)和EGFR未突变组(n=75)。采用卡方检验或t检验比较两组患者临床病理特征及由AI系统测定三期CT图像所得参数(恶性概率预测值、最大径、3D体积、CT值、能量值)。利用二元Logistic回归分析相关变量,建立模型并获得联合预测因子,采用Hosmer-Lemeshow检验对该模型的拟合优度进行检验,对联合预测因子绘制ROC曲线,探讨AI联合临床病理特征预测肺腺癌癌表皮生长因子受体突变的诊断效能。结果两组患者临床特征差异均无统计学意义(P>0.05),AI系统测定三期CT图像所得参数在两组间差异均有统计学意义(P<0.001)。联合预测因子鉴别结节良恶性的ROC曲线下面积AUC为0.957,灵敏度为0.880、特异度为0.880。Hosmer-Lemeshow检验结果表明该模型P>0.10(0.642),拟合效果较好。结论基于人工智能肺结节辅助诊断系统预测肺腺癌表皮生长因子受体突变准确度高,有较好的临床应用价值.

Objective To explore the value of an artificial intelligence-assisted diagnosis system based on CT images in predicting epidermal growth factor receptor mutations in lung adenocarcinoma,and to provide support for clinical diagnosis or decision-making.Methods Retrospective analysis was performed on 143 patients with lung adenocarcinoma confirmed by surgery or needle biopsy in our hospital from 2017 to 2019,and they were divided into EGFR mutation group(n=68)and EGFR non-mutation group(n=75)according to the results of genetic testing.A chi-square test or t-test was used to compare the clinicopathological features of the two groups and the parameters(malignant probability prediction value,maximum diameter,3D volume,CT value,and energy value)obtained from the third-stage CT images determined by the AI system.The Hosmer-Lemeshow test was used to test the good fit of the model.The ROC curve of the combined predictors was drawn to explore the diagnostic efficacy of AI combined with clinicopathological features in predicting epidermal growth factor receptor mutations in lung adenocarcinoma.Binary Logistic regression was used to analyze the relevant variables to establish a model and obtain joint predictive factors.The Hosmer-Lemeshow test was used to test the goodness of fit of the model.The ROC curve was drawn for the joint predictive factors,and the combination of AI and clinical pathological characteristics was used to predict lung glands.Diagnostic efficacy of epidermal growth factor receptor mutations in cancer.Results There was no statistically significant difference in clinical characteristics between the two groups of patients(P>0.05).The parameters obtained from the three-phase CT images measured by the AI system were statistically different between the two groups(P<0.001).The area under the ROC curve of the combined predictor for the identification of benign and malignant nodules was 0.957,the sensitivity was 0.880,and the specificity was 0.880.Hosmer-Lemeshow test results show that the model P>0.10(0.642).The fitting effect is better.Conclusion Based on the artificial intelligence lung nodule diagnosis system,the accuracy of predicting the mutation of an epidermal growth factor receptor in lung adenocarcinoma is high,and it has good clinical application value.