10-02-2017, 12:55 PM
Coronary heart disease (CAD) is the most common type of heart disease that is the leading cause of death worldwide. X-ray angiography is currently the standard gold imaging technique for the diagnosis of CAD. These images often suffer from low quality and presence of noise. Therefore, vessel enlargement and vessel segmentation play an important role in the diagnosis of CAD. In this paper we propose a deep learning approach using convolutional neural networks (CNN) for the detection of vascular regions in angiography images. Initially, an input angiogram is pre-processed to improve its contrast. Subsequently, the image is evaluated using pixel patches and the network determines the regions of the container and the background. A set of 1,040,000 patches is used to train deep CNN. The experimental results on the angiographic images of a data set show that our proposed method has superior performance in the extraction of vascular regions.