基于CNN和局部信息的脑血管图像分割方法

Cerebrovascular Image Segmentation Based on CNN and Local Information

为实现脑血管的分割,提出了一种基于卷积神经网络和局部信息的多模态脑血管图像分割方法。方法对原始脑部CT血管造影图像分别使用高斯滤波和拉普拉斯滤波去除噪声和做锐化处理,和原始图像分别以单个模态作为输入采用卷积神经网络对图像进行血管提取,得到三个模态的分割结果,然后采用加权平均法得到融合结果,最后采用基于局部信息的改进模糊C均值算法对融合结果进行分割得到最终结果。实验结果证明,上述算法比其它算法在脑血管分割上具有更高的有效性。

In order to realize the segmentation of cerebral blood vessels,a multimodal cerebral blood vessel image segmentation method based on convolutional neural network and local information was proposed.In this method,the Gaussian filter and Laplace sharpening filter were used respectively to remove noise and sharpen primitive brain CT angiography images.These results and the original image were used separately with a single modal as the input of convolutional neural networks for blood vessel extraction.Then segmentation results of the three modals were achieved,and the fusion results were obtained using the weighted average method,Finally,the final result was obtained using the improved fuzzy c-means algorithm based on local information to divide the fusion results.Experimental results show that this algorithm is more effective than other algorithms in cerebral vascular segmentation.