基于CT影像评分的人工神经网络模型对肺部良恶性病变的判别价值
Value of artificial neural network model based on pulmonary CT images scores in the diagnosis of lung benign and malignant lesions

目的:应用人工神经网络技术建立基于肺CT影像评分的肺癌辅助诊断模型,探讨其在肺部CT影像良恶性判别中的价值。方法:收集117例肺部病变患者的CT片,由3名有经验的放射科医师提取21项CT影像学特征并量化评分。从总样本中随机抽取73例样本为训练集,余44例为预测集。基于21项CT影像学特征和5项临床参数,应用人工神经网络技术构建肺癌的辅助诊断模型,并与logistic回归模型进行比较。结果:人工神经网络模型对44例样本预测的准确度为90.9%,logistic回归模型为68.2%。结论:与logistic回归模型相比,人工神经网络模型用于判断肺部良恶性病变的准确率高,对于提高肺癌诊断的准确性具有潜在的应用价值。
Aim:To establish an assistant diagnosis model for lung cancer by the artificial neural network(ANN)model on the basis of the scores of lung CT images.Methods:CT images of 117 patients with lung lesions were collected.Twenty-one features of CT images were extracted and then the scores were quantified by 3 experienced radiologists; 73 samples were randomly selected as the training set, and the remaining 44,as the prediction set; finally,an auxiliary diagnostic model for lung cancer was constructed by ANN technology using 21 features of CT images and 5 clinical parameters and its diagnostic value was compared with logistic regression model.Results:The accuracy of ANN model for the prediction set was 90.9%,while that of the logistic regression model was 68.2%.Conclusion:Compared with logistic regression model, the ANN model had higher accuracy in discrimination of lung cancer from lung benign lesions by CT images. It has potential application value in improving the accuracy of lung cancer diagnosis