人工智能作为4-IN-THE-LUNG-RUN肺癌筛查试验首读者的可行性:对阴性误分类和临床转诊率的影响
该研究基于4ITLR试验3678例基线低剂量CT,比较AI(AVIEW-LCS)与放射科医生独立判读的阴性误分类及转诊影响。AI阴性误分类31例(0.8%),放射科医生407例(11.1%);放射科医生首读导致12/102(11.8%)转诊遗漏,高于AI的3/102(2.9%)。结论认为AI阴性误分类显著更少,作为基线筛查首读者有前景。
要点
- 在3678例基线低剂量CT中,AI阴性误分类31例(0.8%),放射科医生407例(11.1%)。
- 放射科医生首读导致12/102(11.8%)转诊遗漏,高于AI的3/102(2.9%)。
- AI作为首读者可独立排除阴性病例,且未大幅增加临床转诊遗漏风险。
核心结论
- AI阴性误分类显著少于放射科医生,作为基线肺癌筛查项目首读者有前景,可独立排除阴性病例而不大幅增加转诊遗漏风险。
研究设计与样本
- 基于欧洲4-IN-THE-LUNG-RUN(4ITLR)试验的3678例基线低剂量CT,由放射科医生和专用AI软件(AVIEW-LCS,v1.1.42.92,Coreline-Soft,首尔,韩国)独立判读,以专家小组为参考标准。
主要发现
- 共识别438例阴性误分类(11.9%),其中AI 31例(0.8%),放射科医生407例(11.1%)。
- 31例AI阴性误分类中,3例分类为阳性,28例为不确定;407例放射科医生阴性误分类中,4例分类为阳性,403例为不确定,其中8例在三个月随访CT后分类为阳性。
- 基线总转诊102例(2.8%);放射科医生首读导致12/102(11.8%)转诊遗漏,高于AI的3/102(2.9%)。
关键数值与推荐
- 阴性误分类定义为存在>100 mm3结节且放射科医生或AI给出阴性分类(仅<100 mm3结节或无结节),以专家小组为参考标准。
- 不确定分类为100-300 mm3,阳性分类为>300 mm3,阳性分类需转诊进行临床检查。
临床怎么用
- 作者认为AI作为基线肺癌筛查项目首读者有前景,可独立排除阴性病例且不大幅增加转诊遗漏风险。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Lung cancer screening (LCS) with low-dose CT (LDCT) reduces lung-cancer-related mortality in high-risk individuals. AI can potentially reduce radiologist workload as first-read-filter by ruling-out negative cases. The feasibility of AI as first reader was evaluated in the European 4-IN-THE-LUNG-RUN (4ITLR) trial, comparing its negative-misclassifications (NMs) to those of radiologists and the impact on referral rates. METHODS: NMs were collected from 3678 baseline LDCTs of the 4ITLR-dataset. LDCTs were read independently by radiologists and dedicated AI software (AVIEW-LCS, v1.1.42.92, Coreline-Soft, Seoul, Korea). A case was designated as NM when nodules > 100 mm3 were present and either radiologist or AI gave a negative-classification (only nodules <100 mm3 or no nodules), with an expert-panel as reference standard. A distinction was made between an indeterminate (100-300mm3), and positive (>300 mm3) classification, warranting referral for clinical-workup. Overall, there were 102 referrals (2.8 %) at baseline. RESULTS: Of the 3678 baseline scans, 438 NMs (11.9 %) were identified (age individuals: 68 (IQR: 64-73) years, 241 men); 31 (0.8 %) by AI and 407 (11.1 %) by radiologists. Among the 31 AI-NMs, 3 were classified positive and 28 indeterminate. Among the 407 radiologist-NMs, 4 were classified positive, and 403 were indeterminate, of which 8 were classified positive after receiving a three-month follow-up CT. Radiologists, as first reader, would have led to 12/102 (11.8 %) missed referrals, higher than the 3/102 (2.9 %) of AI. CONCLUSION: This study showed AI outperforms radiologists with significantly less NMs and therefore shows promise as first reader in a LCS program at baseline, by independently ruling-out negative cases without substantially increasing the risk of missed clinical referrals.
引用
- Walstra ANH, Lancaster HL, Heuvelmans MA, et al. Feasibility of AI as first reader in the 4-IN-THE-LUNG-RUN lung cancer screening trial: impact on negative-misclassifications and clinical referral rate. Eur J Cancer. 2024. PMID: 39754864. doi:10.1016/j.ejca.2024.115214
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。