Convolutional Neural Networks For Segmentation Of Blood Vessel In Medical Images

Medical imaging technics are increasingly crucial for disease diagnosis and treatment, especially the vascular structure imaging technics. Large amounts of statistics show that coronary artery disease currently is one of the leading cause of death worldwide. Simultaneously, coronary artery structure imaging technics play a key role in disease diagnosis and surgery navigation. From the view of clinical application, it is essential to do some fundamental research on the vascular structure imaging technics.For the presence of non-automatic feature extraction from vascular structure images, we study Convolutional Neural Networks for segmentation of blood vessel in medical images. The main works are as follows:Firstly, three-dimensional multi-scale line filter for segmentation of carotid. The three-dimensional multi-scale line filter consists of boundary expansion, bi-Guassian filter and vesselness computation. The experimental results show that the three-dimensional multi-scale line filter segments carotid effectively.Secondly, Convolutional Neural Networks for segmentation of neuron membrane and carotid. The Convolutional Neural Networks mainly consist of convolution layers, max-pooling layers and fully connected layers. We study factors of correct rate of image segmentation with Convolutional Neural Networks for segmenation of neuron membrane and carotid.Thirdly, Convolutional Neural Networks for segmentation of blood vessel in murine hind-limb.The Convolutional Neural Networks and the three-dimensional multi-scale line filter segment blood vessel in murine hind-limb respectively. Compared with the three-dimensional multi-scale line filter, the Convolutional Neural Networks segment blood vessel more effectively.