AI定量参数对良恶性肺结节的诊断价值

Diagnostic Value of AI Quantitative Parameters in Benign and Malignant Pulmonary Nodules

摘要: 目的：探讨AI定量参数对良恶性肺结节诊断价值。方法：回顾性收集2019年12月~2022年1月在延安大学附属医院经手术切术最终病理确诊的孤立性肺结节患者84例，其中肺腺癌62例，肺良性结节22例，利用人工智能软件，在薄层扫描肺窗图像上提取AI定量参数。采用独立样本t检验或非参数检验比较良恶性组患者定量参数差异，进行单因素二元logistic回归分析，并绘制ROC曲线计算曲线下面积(AUC)，敏感度、特异度寻找鉴别良恶性结节的风险因素。结果：良恶性结节的一般临床基线资料差异没有统计学意义。单因素回归分析显示平均CT值、中位数、偏度是良恶性结节预测因素。多因素回归显示平均CT值是肺良恶性结节独立预测指标(OR值 = 0.995，95% CI：0.992~0.999)。ROC曲线下面积AUC = 0.770 (0.615~0.915)，临界值为−165.35，灵敏度62%，特异度87.5%。结论：定量参数平均CT值、中位数、偏度能够鉴别良恶性肺结节，其中平均CT值诊断效能最优。

Abstract: Objective: To evaluate the value of AI quantitative parameters in the diagnosis of benign and ma-lignant pulmonary nodules. Methods: A total of 84 patients with lung solitary pulmonary nodule, including 62 cases of adenocarcinoma and 22 cases of benign pulmonary nodules, were retrospec-tively collected from December 2019 to January 2022 in the affiliated hospital of Yan’an university, using artificial intelligence software, AI quantitative parameters were extracted from thin-slice lung window images. The differences in quantitative parameters between benign and malignant groups were compared by independent-sample t-test or nonparametric test, and univariate binary logistic regression analysis was performed. The area under the curve (AUC) was calculated by plotting the ROC curve, sensitivity and specificity were used to identify the risk factors of benign and malignant nodules. Results: The General Clinical Baseline data for benign and malignant nodules were not sta-tistically significant. Univariate regression analysis showed that mean CT value, median and skew-ness were predictive factors of benign and malignant nodules. Multivariate regression analysis showed that the mean CT value was an independent predictor of benign and malignant pulmonary nodules (OR = 0.995, 95% CI: 0.992~0.999). The area under the ROC curve was AUC = 0.770 (0.615~0.915), the critical value was −165.35, the sensitivity was 62%, the specificity was 87.5%. Conclusion: The mean value, median value and skewness of quantitative parameters could differen-tiate benign and malignant pulmonary nodules, and the mean value of CT was the best.