Research Of Angiogram Image Enhancement

The objective of enhancement for angiography image is to highlight vascular structures and suppress the background and non-vascular structures, moreover make the enhanced result close to the actual vascular structure as much as possible. The processing of angiography image is one of the important parts in the research field of medical image processing and analysis. Angiography images take an important role in the clinical diagnosis based on image-guided computer-assisted surgery and other related fields. It has been a more and more widely used technology that using angiography for a variety diagnostic and interventional treatment of cardiovascular and cerebrovascular diseases. As the distribution of contrast agent is uneven, X-ray exposure unequally and other complex situations exist, the contrast image obtained directly by the related equipment often can not show the local and global structure of the vascular tree clearly, the accuracy of diagnosis is often affected. It’s necessary to research how to highlight the vascular structure of the angiogram for the development of research. And it’s also an important prerequisite for vascular centerline extraction, vascular segmentation, measurement and3D visualization. For the above reasons, the enhancement of angiography image has been a research hot in medical image processing and analysis.In recent decades, there have been varieties of algorithm for the enhancement of angiography image, but they still can not meet the actual need completely. The reason is quite complex, such as the sensitive to background noise, boundary shrinking introduced by the Gaussian filter, non-adaptive parameters, poor identified result for small blood vessels enhanced, difficult to identify the bifurcation point and the node, poor computing speed, and rough enhanced vascular surface, and so on. These determine that a perfect common enhancement method could not be obtained and only a reasonable choice could be acquired by considering indicators of the precision, speed and robustness for specific problems and specific needs.At present, the most classic and the latest the method used for angiography image enhancement of is as follows:(1) the method based on Hessian matrix, such as Frangi filter, Sato filter, Lorenz filter;(2) the method based on diffusion equation, such as Regularized Perona-Malik, Edge enhancing diffusion, Coherence enhancing diffusion;(3) strain energy filter (4) polar neighborhood intensity profile filterVessel enhancement diffusion algorithm(VED) is a combination of Frangi filter based on Hessian matrix and nonlinear anisotropic diffusion equation, and Frangi vesselness measurement function is transformed, a smooth constraints is added, so that function is continuous, can be directly used to build diffusion tensor. Two groups experimental result are compared and analyzed for vessel enhancement diffusion algorithm and several other vessel enhancement algorithm, visualization shows under the maximum density projection and volume rendering respectively. The experiment shows that vessel enhancement algorithm can well strengthen vessel structure, while preserving blood vessel edge information and it is also very good to enhance small blood vessels structure, make blood vessel structure become smooth and clear, improve the visual display effect.In reality, because medical vessel image have a big data quantity (many are three dimensional) and rich grey level. When solute maximum vessel function response and compute nonlinear anisotropic diffusion partial differential equation under using a multiscale approach, which have a large computational burden, computing time often above several hours, it seriously influence this algorithm clinical practice. Actual, usually reduced choice scales and iterative numbers suitably, not only will affect the algorithm the accuracy, but simultaneously will limit the computing time promotion.Compute unified device architecture (CUDA) has freed general purpose graphic processing unit(GPGPU) from the graphics fixed pipeline and high level shading language, allowing the design and implementation of SIMD(single instruction and multiple data) parallel algorithms on a much more simple way than previous method based on texture rendering. The GPGPU computing architecture provides a similar C language development environment, allows designers direct use the GPU computing resources through C and CUDA programming language to reduce the development complexity and enhance the development efficiency.The core concept of CUDA was first studied, by taking use of the independence of image pixel and concurrence of partial differential equation, then the vessel enhancing diffusion algorithm was redesigned into parallel model, in implementations, the algorithm is optimized combined with the impact on the computing performance of the CUDA global memory access, thread shared memory allocation block size and allocation of shared memory, comparing with the traditional CPU serial algorithm, this method not only maintains the same good performance of vessel enhancement but compute speed of image data has been significantly improved.