目的 探讨基于人工智能(AI)的影像学模型,结合计算机断层扫描(CT)形态学特征,在鉴别肺结节良恶性及预测其术后病理结果方面的价值。方法 本研究根据术后病理结果,将患者分为良性组与恶性组。收集患者的临床资料与CT影像学特征,并采用AI对CT图像进行分析。通过单因素与多因素logistic回归分析,筛选与病理结果独立相关的因素,并筛选影响AI将良性结节误判为恶性结节的因素,评估基于AI构建的模型的诊断效能。结果 单因素分析表明,患者性别、多项CT特征(包括影像形态、平均CT值、血管征、毛刺征、空泡征、胸膜凹陷征、钙化、CT影像特征)及AI分析结果与病理结果显著相关(P<0.05),CT影像形态、结节直径、钙化、平均CT值与AI模型误判相关(P<0.05)。基于AI构建的临床预测模型(AUC=0.8837)较由多因素分析结果构建的临床预测模型(AUC=0.8398),有更好的诊断效能。结论 基于AI的模型分析以及关键的CT形态学特征是区分肺结节良恶性的独立影响因素,其预测效能优于单一的传统临床指标,为临床筛检高危肺结节、实现肺癌早期精准干预提供了有力工具。
Abstract
Objective To explore the value of an artificial intelligence (AI)-based imaging model combined with morphological features of computed tomography (CT) in differentiating the benign pulmonary nodules from the malignant ones and predicting the postoperative pathological outcomes. Methods Based on the postoperative pathological results, the selected patients were divided into a benign group and a malignant group. The clinical data and CT imaging features of the patients were collected, and AI was used to analyze the CT images. Univariate and multivariate logistic regression analyses were employed to identify factors independently associated with pathological outcomes, as well as factors associated with AI model misdiagnosis (i.e., false-positive classification of benign nodules). These analyses served to evaluate the diagnostic performance of the AI model. Results Univariate analysis indicated that patient gender, multiple CT features (including morphological appearance, mean CT value, vascular sign, spiculation, cavitation, pleural retraction, calcification and CT imaging features), and AI analysis results were significantly correlated with the pathological outcomes (all P<0.05). Separately, morphological appearance, nodule diameter, calcification, and mean CT value showed significant association with AI model misclassification (all P<0.05). The AI-based clinical prediction model (AUC=0.8837) had better diagnostic efficacy than the clinical prediction model constructed from multivariate analysis results (AUC=0.839 8). Conclusions The analysis of AI-based model and key CT morphological features are independent influencing factors in differentiating benign and malignant pulmonary nodules. Its predictive efficacy is better than a single traditional clinical indicator, providing a powerful tool for clinical screening of high-risk pulmonary nodules and achieving early precise intervention for lung cancer.
关键词
AI /
肺癌 /
计算机断层扫描 /
肺结节 /
人工智能
Key words
AI /
lung cancer /
CT /
pulmonary nodule /
artificial intelligence
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Couraud S, Ferretti G, Milleron B, et al. Diagnostic and Interventional Imaging Intergroupe francophone de cancérologie thoracique, société de pneumologie de langue française, and société d’imagerie thoracique statement paper on lung cancer screening[M].Parris Elsevier Masson, 2021,102(4):199-211.
[2] Barta J A, Powell C A, Wisnivesky J P.Global epidemiology of lung cancer[J]. Ann Glob Health,2019,85(1):8.
[3] Aberle D R, Adams A M, Berg C D, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening[J]. N Engl J Med,2011,365(5):395-409.
[4] de Koning H J, van der Aalst C M, de Jong P A, et al. Reduced lung-cancer mortality with volume CT screening in a randomized trial[J]. N Engl J Med, 2020,382(6):503-513.
[5] El Ayachy R, Giraud N, Giraud P, et al.The role of radiomics in lung cancer:from screening to treatment and follow-up[J]. Front Oncol, 2021,11:603595.
[6] Lancaster H L, Zheng S, Aleshina O O, et al.Outstanding negative prediction performance of solid pulmonary nodule volume AI for ultra-LDCT baseline lung cancer screening risk stratification[J]. Lung Cancer,2022,165:133-140.
[7] Chassagnon G, de Margerie-Mellon C, Vakalopoulou M, et al.Artificial intelligence in lung cancer:current applications and perspectives[J]. Jpn J Radiol,2023,41(3):235-244.
[8] Hu F, Huang H, Jiang Y, et al.Discriminating invasive adenocarcinoma among lung pure ground-glass nodules:a multi-parameter prediction model[J]. J Thorac Dis,2021,13(9):5383-5394.
[9] Christensen J, Prosper A E, Wu C C, et al.ACR Lung-RADS v2022:assessment categories and management recommendations[J]. Chest,2024,165(3):738-753.
[10] Oken M M, Hocking W G, Kvale P A, et al.Screening by chest radiograph and lung cancer mortality:the prostate, lung, colorectal, and ovarian (PLCO) randomized trial[J]. JAMA,2011,306(17):1865-1873.
[11] McDonald R J, Schwartz K M, Eckel L J, et al.The effects of changes in utilization and technological advancements of cross-sectional imaging on radiologist workload[J]. Acad Radiol, 2015,22(9):1191-1198.
[12] Hanna T N, Lamoureux C, Krupinski E A, et al.Effect of shift, schedule, and volume on interpretive accuracy:a retrospective analysis of 2.9 million radiologic examinations[J]. Radiology, 2018,287(1):205-212.
[13] Baldwin D R, Gustafson J, Pickup L, et al.External validation of a convolutional neural network artificial intelligence tool to predict malignancy in pulmonary nodules[J]. Thorax,2020,75(4):306-312.
[14] Radiology ACo. Lung CT screening reporting & data system Lung-RADS Version 1.1[M].Reston:American College of Radiology,2020.
[15] Cai Y, Chen T, Zhang S, et al.Correlation exploration among CT imaging, pathology and genotype of pulmonary ground-glass opacity[J]. J Cell Mol Med,2023,27(14):2021-2031.
[16] Tanner N T, Porter A, Gould M K, et al.Physician assessment of pretest probability of malignancy and adherence with guidelines for pulmonary nodule evaluation[J]. Chest, 2017,152(2):263-270.
[17] Fox A H, Tanner N T.Approaches to lung nodule risk assessment:clinician intuition versus prediction models[J]. J Thorac Dis,2020,12(6):3296-3302.
[18] Ardila D, Kiraly A P, Bharadwaj S, et al.End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography[J]. Nat Med,2019,25(6):954-961.
[19] Sundaram V, Gould M K, Nair V S.A comparison of the pancan model and Lung-RADS to assess cancer probability among people with screening-detected, solid lung nodules[J] Chest,2021,159(3):1273-1282.
PDF(2163 KB)
