Diagnostic efficacy of artificial intelligence in distinguishing benign and malignant pulmonary nodules: based on different who pathological classifications of lung cancer
人工智能对肺结节良恶性的诊断效能：基于不同WHO肺癌病理分类

摘要

目的 对比2015版与2021版WHO肺部肿瘤病理诊断标准，探究智能影像识别技术在肺结节良恶性鉴别中的临床应用效能。方法 选取本院2023年3月-2025年3月100例肺结节患者为研究对象，以术后病理结果为诊断金标准，结合两版病理规范界定腺体前驱病变，系统评估智能影像诊断技术的灵敏度、特异度、诊断一致性及受试者工作特征曲线下面积（AUC）值。结果 共检出108枚肺结节，含18枚腺体前驱病变。智能诊断技术适配旧版标准的诊断效能更优，新版标准下整体诊断精准度有所下降；结节CT均值、影像熵值为恶性结节的独立影响因素。结论 WHO病理分型标准更新会影响智能影像的诊断效果，需结合最新病理规范优化智能诊断模型，提升肺结节鉴别诊断的精准度。

Abstract

Objective To compare the 2015 and 2021 editions of the WHO lung tumor pathological diagnosis criteria, and explore the clinical application efficacy of intelligent image recognition technology in the differentiation of benign and malignant pulmonary nodules.
Methods A total of 100 patients with pulmonary nodules from our hospital from March 2023 to March 2025 were selected as the research subjects. The postoperative pathological results were used as the diagnostic gold standard, and the adenoma precursor lesions were defined according to the two editions of pathological norms. The sensitivity, specificity, diagnostic consistency, and Area under the receiver operating characteristic curve (AUC) value of the intelligent image diagnosis technology were systematically evaluated.
Results A total of 108 pulmonary nodules were detected, including 18 adenoma precursor lesions. The diagnostic efficacy of the intelligent diagnosis technology adapted to the old standard was better, while the overall diagnostic accuracy decreased under the new standard; the mean value of CT of the nodules and the image entropy value were independent influencing factors for malignant nodules.
Conclusion   The update of the WHO pathological classification criteria will affect the diagnostic effect of intelligent imaging. It is necessary to optimize the intelligent diagnosis model by combining the latest pathological norms to improve the accuracy of pulmonary nodule differential diagnosis.