S1 Appendix -

Lung cancer is a common malignant tumor disease with high clinical disability and death rates. Currently, lung cancer diagnosis mainly relies on manual pathology section analysis, but the low efficiency and subjective nature of manual film reading can lead to certain misdiagnoses and omissions. With the continuous development of science and technology, artificial intelligence (AI) has been gradually applied to imaging diagnosis. Although there are reports on AI-assisted lung cancer diagnosis, there are still problems such as small sample size and untimely data updates. Therefore, in this study, a large amount of recent data was included, and meta-analysis was used to evaluate the value of AI for lung cancer diagnosis. With the help of STATA16.0, the value of AI-assisted lung cancer diagnosis was assessed by specificity, sensitivity, negative likelihood ratio, positive likelihood ratio, diagnostic ratio, and plotting the working characteristic curves of subjects. Meta-regression and subgroup analysis were used to investigate the value of AI-assisted lung cancer diagnosis. The results of the meta-analysis showed that the combined sensitivity of the AI-aided diagnosis system for lung cancer diagnosis was 0.87 [95% CI (0.82, 0.90)], specificity was 0.87 [95% CI (0.82, 0.91)] (CI stands for confidence interval.), the missed diagnosis rate was 13%, the misdiagnosis rate was 13%, the positive likelihood ratio was 6.5 [95% CI (4.6, 9.3)], the negative likelihood ratio was 0.15 [95% CI (0.11, 0.21)], a diagnostic ratio of 43 [95% CI (24, 76)] and a sum of area under the combined subject operating characteristic (SROC) curve of 0.93 [95% CI (0.91, 0.95)]. Based on the results, the AI-assisted diagnostic system for CT (Computerized Tomography), imaging has considerable diagnostic accuracy for lung cancer diagnosis, which is of significant value for lung cancer diagnosis and has greater feasibility of realizing the extension application in the field of clinical diagnosis.

肺癌是一类临床致残率与病死率均较高的常见恶性肿瘤。当前肺癌诊断主要依赖人工病理切片分析，但人工阅片效率低下且主观性较强，易造成漏诊与误诊。随着科技的持续进步，人工智能（AI）已逐步应用于影像诊断领域。尽管已有关于AI辅助肺癌诊断的相关报道，但仍存在样本量偏小、数据更新不及时等问题。因此本研究纳入大量近期数据，采用荟萃分析（meta-analysis）评估AI用于肺癌诊断的价值。本研究借助STATA16.0软件，通过特异度、敏感度、阴性似然比、阳性似然比、诊断比值比，并绘制受试者工作特征（ROC）曲线，对AI辅助肺癌诊断的价值进行评估；同时采用Meta回归与亚组分析，进一步探究其应用价值。荟萃分析结果显示，AI辅助诊断系统用于肺癌诊断的合并敏感度为0.87[95%CI(0.82, 0.90)]，合并特异度为0.87[95%CI(0.82, 0.91)]（CI即置信区间），漏诊率与误诊率均为13%，阳性似然比为6.5[95%CI(4.6, 9.3)]，阴性似然比为0.15[95%CI(0.11, 0.21)]，诊断比值比为43[95%CI(24, 76)]，综合受试者工作特征（SROC）曲线下面积为0.93[95%CI(0.91, 0.95)]。基于上述研究结果，AI辅助计算机断层扫描（CT）影像诊断系统在肺癌诊断中具备可观的诊断精准度，对肺癌诊断具有重要临床价值，在临床诊断领域推广应用的可行性较高。