基于深度学习算法的CT自动诊断系统在NSCLC患者TNM分期中的建立与应用  

Establishment and Application of a CT Automatic Diagnosis System based on Deep Learning Algorithm in TNM staging of NSCLC Patients

作　　者：王敏[1] 阮振平 沈霁 WANG Min;RUAN Zhen-ping;SHEN Ji(Radiology Department of the Second Hospital of Anhui Medical University/Medical Imaging Research Center of Anhui Medical University,Hefei 230601,Anhui Province,China)

机构地区：[1]安徽医科大学第二附属医院放射科/安徽医科大学医学影像研究中心,安徽合肥230601

出　　处：《中国CT和MRI杂志》2025年第3期67-70,共4页Chinese Journal of CT and MRI

摘　　要：目的建立基于深度学习算法的计算机断层扫描(CT)自动诊断系统,并探讨其在非小细胞肺癌(NSCLC)患者肿瘤淋巴结转移(TNM)分期中的应用。方法回顾性分析本单位2020年1月~2022年6月收治的245例NSCLC患者的临床资料作为训练集,提取CT检查特征信息和TNM分期数据,基于深度学习算法构建CT自动诊断系统。另选取本单位2022年7月~2023年6月收治的102例NSCLC患者作为验证集,经该系统诊断TNM分期,将病理学检查结果作为“金标准”,评价该系统对TNM分期的诊断价值。结果基于深度学习算法构建了CT自动诊断系统,该系统诊断各期的灵敏度、特异度、准确度与约登指数均高于常规CT诊断;基于深度学习算法的CT自动诊断系统诊断TNM分期与病理学诊断一致性高(Kappa=0.846,P<0.001),常规CT诊断TNM分期与病理学诊断也有一致性(Kappa=0.721,P<0.001)。结论基于深度学习算法的CT自动诊断系统诊断TNM分期优于常规CT。Objective To establish a computer tomography(CT)automatic diagnosis system based on deep learning algorithms and explore its application in staging of tumor node metastasis(TNM)in nonsmall cell lung cancer(NSCLC)patients.Methods The clinical data of 245 NSCLC patients admitted to our department from January 2020 to June 2022 were retrospectively analyzed as a training set,which was extracted CT examination feature information and TNM staging data,and a CT automatic diagnosis system based on deep learning algorithms was constructed.102 NSCLC patients admitted to our department from July 2022 to June 2023 were selected as the validation set.TNM staging was diagnosed using this system,and pathological examination results were used as the"gold standard"to evaluate the diagnostic value of this system for TNM staging.Results A CT automatic diagnosis system was constructed based on deep learning algorithms.The sensitivity,specificity,accuracy,and Youden index of the system for each stage of diagnosis were higher than those of conventional CT diagnosis.The CT automatic diagnosis system based on deep learning algorithm had high consistency between TNM staging and pathological diagnosis(Kappa=0.846,P<0.001),and conventional CT diagnosis of TNM staging and pathological diagnosis also had consistency(Kappa=0.721,P<0.001).Conclusion The CT automatic diagnosis system based on deep learning algorithm is superior to conventional CT in diagnosing TNM staging.

关 键 词：深度学习算法 计算机断层扫描 自动诊断系统 非小细胞肺癌 分期 

分 类 号：R734.2[医药卫生—肿瘤]