Our intent in this project has been to improve the small vessel visibility in MIP images of 3D TOF MRA studies. During the first 3 years we tested the effectiveness of a non-linear vessel enhancing filter applied to the original 3D TOF MRA image data. In trying to understand the strength of the MIP algorithm in vessel display and to understanding the properties of vessels, we noticed the striking difference between the nature of """"""""vessels"""""""" and background in the MIP depth buffer (the matrix of z-locations of points projected in the MIP). Vessels are predominantly smooth and connected while background is """"""""rough"""""""". In trying to utilize this property to enhance the probability of vessel pixels, we have developed an algorithm which nearly completed extracts vessel voxels from the 3D MRA data, and excludes nearly all background. Using these extracted voxels, we have been able to generate X-ray like projection images of the 3D MRA image data which inherently contained more information than the MIP, leading to a striking improvement in vessel appearance. Nearly every vessel seen in the original MAP and some not seen are visible in these reproject images with the exact appearance of a digital subtraction X-ray angiogram (DSA). In the next funding period we will perform a large series of tasks designed at refining this new depth buffer segmentation (DBS) algorithm. After refining the algorithm, we will test the extent to which the DBS algorithm improves the accuracy and efficiency of detection and management of a variety of intracranial pathologies such as aneurysms and vasculitis. We will also study the application to other vascular systems. Because the algorithm extracts the image coordinates of the voxels that make up the vessels for which segments are visible in the MIP, we will develop algorithms which convert these lists of segmented voxels to a cubic spline representation, where the anatomic labels of the vessels are included. We will study methods to use this knowledge of the patients vascular anatomy to develop techniques to assist in the optimized presentation of the MRA information for improved diagnostic accuracy and efficiency. We will also test the application of the 3D vessel structure obtained from the DBS algorithm to improve surgical procedure planning and other applications. We believe that an improved version of this exciting new algorithm will improve the efficiency and accuracy of MRA in general. Experiments to characterize, improve and clinically evaluate this algorithm are described in this proposal.
