基于深度学习的CAD系统结合低剂量CT在孤立性肺结节良恶性鉴别中的价值OA
Value of Deep Learning-Based CAD System Combined with Low-dose CT in Differential Diagnosis of Benign and Malignant Solitary Pulmonary Nodules
目的:探讨基于深度学习的计算机辅助检测(CAD)系统结合低剂量 CT 在孤立性肺结节良恶性鉴别中的价值,为临床早期肺癌筛查及精准诊断提供参考.方法:回顾性选取 2023 年 1 月—2025 年 6 月宜春市人民医院收治的 82 例孤立性肺结节患者作为研究对象,所有患者均接受低剂量 CT 扫描并经病理检查确诊.收集患者的低剂量CT影像数据及临床资料,以病理检查结果为金标准,对比基于深度学习的 CAD 系统与传统影像诊断(主治医师阅片)对孤立性肺结节良恶性鉴别诊断的效能,并对比二者在不同大小、位置结节中的良恶性鉴别诊断准确率.结果:经病理诊断,82 例肺结节患者中,良性62 例,恶性 20 例.以病理结果为金标准,基于深度学习的 CAD 系统在肺结节良恶性鉴别中的敏感度、特异度、阳性预测值、阴性预测值均高于传统影像诊断,但差异无统计学意义(P>0.05),两种检测方法准确率比较差异有统计学意义(P<0.05).按结节大小分,<5 mm结节34个,5～10 mm结节46个,>10 mm结节 2 个;按结节位置分,肺上叶结节 46 个,肺中叶结节 6 个,肺下叶结节 30 个;两种诊断方式在不同位置结节中的鉴别准确率差异无统计学意义(P>0.05);5～10 mm 结节中,基于深度学习的 CAD 系统诊断准确率高于传统影像诊断(P<0.05),其他大小结节两种诊断方法准确率比较,差异无统计学意义(P>0.05).结论:基于深度学习的 CAD 系统结合低剂量 CT,鉴别孤立性肺结节良恶性整体效能优于传统影像诊断,尤其对 5～10 mm结节优势显著.
Objective:To explore the value of deep learning-based computer-aided design(CAD)system combined with low-dose CT in differential diagnosis of benign and malignant solitary pulmonary nodules,to provide reference for early clinical screening and accurate diagnosis of lung cancer.Method:Totally 82 patients with solitary pulmonary nodules admitted to Yichun People's Hospital from January 2023 to June 2025 were retrospectively selected.All patients received low-dose CT scan and were confirmed by pathological examination.Their low-dose CT imaging data and clinical data were collected.With pathological results as the gold standard,the efficacy of deep learning-based CAD system in differential diagnosis of benign and malignant solitary pulmonary nodules was compared with that of traditional imaging diagnosis(film reading conducted by the attending doctor).The accuracy rates in differential diagnosis of nodules in different sizes,and at different sites was compared between the two methods.Result:Pathological diagnosis confirmed 62 benign cases and 20 malignant cases among the 82 patients.With pathological results as the gold standard,deep learning-based CAD system demonstrated higher sensitivity,specificity,positive and negative predictive values in differential diagnosis of benign and malignant pulmonary nodules compared to traditional imaging diagnosis,with no statistically significant differences(P>0.05).However,there was a statistically significant difference in accuracy rate between the two methods(P<0.05).From the perspective of size,there were 34 nodules smaller than 5 mm,46 nodules ranging from 5-10 mm,and 2 nodules greater than 10 mm.In terms of location,there were 46 nodules in the upper lobe,6 in the middle lobe,and 30 in the lower lobe.There was no statistically significant difference in the diagnostic accuracy rate between the two methods for nodules at different sites(P>0.05).For nodules sized 5-10 mm,deep learning-based CAD system demonstrated significantly higher diagnostic accuracy rate than traditional imaging diagnosis(P<0.05).For nodules in other sizes,no statistically significant difference in diagnostic accuracy rate was observed(P>0.05).Conclusion:Deep learning-based CAD system combined with low-dose CT is more efficient than traditional imaging diagnosis for identifying benign and malignant solitary pulmonary nodules,particularly for nodules sized 5-10 mm.
肖宇;黄仕华;吴鸿波;张燕香;易玉涛
宜春市人民医院放射科 江西 宜春 336000宜春市人民医院放射科 江西 宜春 336000宜春市人民医院放射科 江西 宜春 336000宜春市人民医院放射科 江西 宜春 336000宜春市人民医院放射科 江西 宜春 336000
医药卫生
低剂量CT孤立性肺结节深度学习计算机辅助检测系统
Low-dose CTSolitary pulmonary noduleDeep learningComputer-aided detection system
《中国医学创新》 2026 (15)
108-112,5
江西省卫生健康委科技计划项目(202511183)
评论