摘要
目的探讨神经网络对高分瓣率CT上孤立性肺结节(solitary pulmonary nodule,SPN)的诊断性能。方法共收集孤立性肺结节145例,其中原发性肺癌86例,结核球18例,炎性结节29例,良性肿瘤12例,均经手术病理或CT导向经皮肺穿刺活检病理组织学证实以及2年以上临床治疗追踪确诊。观察5项临床指标和10项CT影像学指标,并对定性指标进行量化。从145个样本中随机选择70%左右的样本(103例)作为训练样本,建立BP神经网络诊断模型和Logistic回归模型。比较两种模型对所有样本的诊断正确率和ROC曲线下面积。结果BP神经网络的诊断正确率为98.6%,高于Logistic回归模型的正确率88.3%(P=0.0007);ROC曲线下面积分别为0.997±0.004和0.959±0.016,差异有统计学意义(P=0.009)。结论利用神经网络诊断孤立性肺结节有良好的诊断性能,值得进一步探讨。
Objective To discuss the performance of artificial neural networks （NN） in diagnosis of solitary pulmonary nodule （SPN） on HRCT images. Method 145 cases of SPN, including 86 cases of pulmonary carcinoma, 18 cases of tuberculomas, 29 cases of infiaannatory nodule and 12 cases of benign tumor, were collected, which were all confirmed by pathology or biopsy over-two-year clinical treatment. Five clinical parameters and 10 radiological findings were observed and quantified for qualitative characteristics. About 70 percent of all cases （up to 103 cases） were selected randomly to form training samples set, on which BP neural networks and Logistic regression model were built. The diagnostic consistent rates and areas under ROC of the two models were then compared. Result The total consistent rate of BP NN, which is 98.6%, is greater than that of Logistic model, which is 88.3% （ P = 0.0007）. Areas under ROC curve are 0.997 ± 0.004 and 0.959 ± 0.016 respectively, and the difference between the two models is significant statistically （ P = 0.009）. Conclusion NN showed high performance in diagnosis of SPN on HRCT images. It was worthy of further study.
出处
《北京生物医学工程》
2005年第6期436-439,共4页
Beijing Biomedical Engineering
基金
首都医科大学基础临床合作项目资助
关键词
孤立性肺结节
神经网络
高分辨率
CT
诊断
solitary pulmonary nodule （SPN）
neural networks （NN）
high resolution CT （HRCT）
diagnosis
