人工智能辅助５１２层螺旋ＣＴ三维扫描对肺恶性结节诊断价值

廖忠剑 王肇平 刘艳平 邓星星 黄颖文

[摘要] 目的 探讨人工智能（AI）辅助512层螺旋CT三维扫描对肺恶性结节诊断价值。 方法 选取2019年6月至2020年12月我院109例肺恶性结节患者，均接受512层螺旋CT三维扫描，分别经AI阅片及人工阅片。统计分析人工阅片及AI阅片用时、人工阅片及AI阅片对不同直径、不同部位肺恶性结节检出情况。 结果 AI阅片用时为（0.51±0.13）min/例，短于人工阅片的（2.98±1.01）min/例，差异有统计学意义（P<0.05）;AI阅片对直径>10 mm的肺恶性结节检出率（98.41%）与人工阅片（95.24%）间比较，差异无统计学意义（P>0.05），AI阅片对直径<5 mm、直径5～10 mm的肺恶性结节检出率为（97.35%、96.63%），高于人工阅片的（80.53%、80.90%），两组比较，差异有统计学意义（P<0.05）;人工阅片对肺门、中间区域、外围区域、胸膜下肺恶性结节检出率为100.00%、98.11%、99.35%、90.00%，高于人工阅片的33.33%、79.25%、92.16%、74.00%，差异有统计学意义（P<0.05）。 结论 AI辅助512层螺旋CT三維扫描在肺恶性结节中具有较高应用价值，可提升不同直径和位置结节检出率，缩短阅片用时，且利于减小影像科医师工作负担。

[关键词] 人工智能;512层螺旋CT;肺恶性结节;诊断价值

[中图分类号] R445.2          [文献标识码] B          [文章编号] 1673-9701（2021）31-0135-03

[Abstract] Objective To investigate the diagnostic value of artificial intelligence-assisted（AI-assisted）512-slice spiral computed tomography（CT）3D scan for pulmonary malignant nodules（PMN）. Methods A total of 109 patients with PMN in our hospital from June 2019 to December 2020 were selected， all of whom underwent 512-slice spiral CT 3D scans. The results were read by AI and manually， respectively. The time spent for manual and AI reading， and the detection rate of PMN of different diameters and at different sites by manual and AI reading were statistically analyzed. Results The time spent for AI reading was（0.51±0.13）minutes/case， which was shorter than that for manual reading（2.98±1.01）minutes/case， and the difference between the two groups was statistically significant（P<0.05）.There was no significant difference in the detection rates of PMN of diameter >10 mm between the AI reading group（98.41%）and the manual reading group（95.24%，P>0.05）. The detection rates of PMN of diameters<5 mm and 5-10 mm in the AI reading group（97.35% and 96.63%）were higher than those in the manual reading group（80.53% and 80.90%）， with statistically significant differences（P<0.05）.The detection rates of PMN at the pulmonary hilum， middle region，peripheral region， and subpleura in the AI reading group（100.00%，98.11%，99.35%，90.00%）were higher than those in the manual reading group（33.33%，79.25%， 92.16%，74.00%）， with statistically significant differences（P<0.05）. Conclusion AI-assisted 512-slice spiral CT 3D scan has a high application value in the diagnosis of PMN， which can improve the detection rate of nodules of different diameters and at different sites， shorten the time spent on film reading， and help to reduce the workload of imaging physicians.

[Key words] Artificial intelligence;512-slice spiral CT;Pulmonary malignant nodules;Diagnostic value

肺恶性结节为临床多发恶性肿瘤疾病，我国为肺恶性结节多发地区，近年来，受不良生活习惯养成、环境污染、抽烟等诸多因素影响，其发病率持续增高，对患者生活质量及身心健康构成了极大威胁[1-2]。同时，肺恶性结节具有较高病死风险，整体5年生存率不足15%，而ⅠA期患者经手术切除后长期生存率则可达80%，故早期对疾病进行明确诊断对指导临床及早采取对应治疗方案以延长患者生存期限具有重要意义[3-4]。……