摘要
目的 探究人工智能(AI)在改善肺磨玻璃结节(GGN)CT图像重建质量中的应用价值。方法 选取2022年3月至2024年3月四川省自贡市第四人民医院172例GGN患者,均行胸部CT检查,分别进行人工后处理和AI后处理。比较人工后处理和AI后处理的GGN检出率、形态学特征检出率及重建图像质量,以病理诊断结果为“金标准”,比较2种方法鉴别诊断GGN良恶性与病理诊断结果的一致性,比较人工后处理和AI后处理对GGN良恶性的鉴别诊断价值。结果 AI后处理的GGN检出率为90.70%,高于人工后处理的80.81%(P<0.05);AI后处理测定的结节直径及毛刺征、胸膜粘连征、分叶征、空洞、钙化检出率与人工后处理比较,差异无统计学意义(P>0.05);AI后处理的整体重建图像质量、GGN图像质量评分及信噪比(SNR)均高于人工后处理(P<0.05);AI后处理鉴别诊断GGN良恶性结果与病理诊断结果的一致性Kappa值为0.666(95%CI:0.519~0.813),人工后处理鉴别诊断GGN良恶性结果与病理诊断结果的一致性Kappa值为0.569(95%CI:0.420~0.718);AI后处理鉴别诊断GGN良恶性的敏感度为92.31%,高于人工后处理的80.77%(P<0.05)。结论 AI能明显改善GGN患者的CT图像重建质量,且能提高GGN检出率、鉴别诊断GGN良恶性的敏感度。
Objective To explore the application value of artificial intelligence(AI)in improving the quality of CT image reconstruction of pulmonary ground glass nodules(GGN).Methods A total of 172 GGN patients in Zigong Fourth People's Hospital from March 2022 to March 2024 were selected,all of whom underwent chest CT examination,manual post-processing and AI post-processing.The GGN detection rate,morphological feature detection rate and reconstructed image quality,and AI post-processing were compared.Pathological diagnosis results were taken as the"gold standard"to analyze the consistency of the two methods in the differential diagnosis of benign and malignant GGN,and the value of manual post-processing and AI post-processing in the differential diagnosis of benign and malignant GGN was compared.Results The detection rate of GGN by AI post-processing was significantly higher by manual post-processing(90.70%vs 80.81%,P<0.05).There were no significant differences in the nodule diameter,burr sign,cavity sign,foliation sign,pleural adhesion sign and calcification detected by AI post-processing and manual post-processing(P>0.05).The overall reconstructed image quality,GGN image quality score and signal-to-noise ratio(SNR)of AI post-processing were significantly higher than those of manual post-processing(P<0.05).The Kappa value of AI post-processing in the differential diagnosis of benign and malignant GGN was 0.666(95%CI:0.519-0.813),and that of manual post-processing was 0.569(95%CI:0.420~0.718).The sensitivity of AI post-processing was 92.31%,which was higher than that of manual post-processing(80.77%)(P<0.05).Conclusion AI can significantly improve the CT image reconstruction quality of GGN,and enhance the detection rate of GGN and the sensitivity of differential diagnosis of benign and malignant GGN.
作者
谢刚
姚本波
李燕
XIE Gang;YAO Benbo;LI Yan(Department of Radiology,Zigong Fourth People's Hospital,Sichuan,Zigong 643000,China;不详)
出处
《河北医药》
2025年第7期1175-1178,共4页
Hebei Medical Journal
基金
四川省医学会医学科研(青年创新)课题计划(编号:Q22087)。
关键词
肺磨玻璃结节
人工智能
CT
重建
图像质量
检出率
pulmonary ground glass nodules
artificial intelligence
CT
rebuilding
image quality
detection rate
