基于改进的VGG-16卷积神经网络的肺结节检测

【摘要】针对肺结节特征复杂、人工提取特征困难的问题，提出基于改进的VGG-16卷积神经网络的肺结节检测模型。首先采用阈值分割与处理最大连通区域后的图像进行掩模运算，得到肺实质部分。然后通过Regionprops标记每个连通区域序号分割出所有疑似结节；采用核函数极限学习机而不是Softmax函数作为VGG-16结构中的分类器。最后利用改进后的VGG-16模型去除假阳性结节，完成对肺结节检测。在LIDC-IDRI数据集上进行的实验表明改进后的模型能达到92.56%的准确率和94.44%的高敏感度。该模型可用于辅助医生进行肺结节诊断，具有一定的临床应用价值。

Detection of pulmonary nodules based on improved VGG-16 convolution neural network

Abstract: Aiming at the complex features of pulmonary nodules and the difficulties of extracting features manually, a pulmonary nodule detection model based on improved VGG-16 convolution neural network is proposed. Firstly, the lung parenchyma is obtained by threshold segmentation and mask operation after processing the image of the maximum connected area. Then, the serial number of each connected area is labeled by Regionprops for obtaining all suspected nodules. Kernel extreme learning machine instead of Softmax function is taken as classifier in VGG-16 architecture. Finally, the improved VGG-16 model is used to remove false positive nodules and complete the detection of pulmonary nodules. The proposed method is tested on LIDC-IDRI dataset, and the results showed that the improved model can achieve an accuracy of 92.56% and a sensitivity up to 94.44%. The proposed model can be used to assist doctors in the diagnosis of pulmonary nodules, and has a certain clinical value.