BIOMEDICAL IMAGE PROCESSING

点击翻译按钮获取中文摘要

This project uses sophisticated image processing techniques to analyze biomedical images. The goal is to establish collaborations with biomedical experts who require new algorithms and possibly new hardware capability to solve difficult imaging problems. Typically, complex new mathematical algorithms as well as new combinations of existing algorithms are utilized. We attempt to integrate the best computer platform for each problem with the desired goal of the project, using such diverse computers as an Apple MacIntosh, a DEC VAX or Alpha, a SUN workstation, or an Intel iPSC/860 supercomputer. Two current projects include ophthalmic image acquisition and analysis and ultrasound image analysis. In the first project, research has been ongoing between the National Eye Institute (NEI) and DCRT in the area of computerizing instrumentation and the automatic analysis of anterior eye segment images. The goal in development of computer-based NEI instrumentation is two fold: (a) to provide accurate and reproducible numerical information (b) to develop image analysis in a user-friendly and systematic way. This year we devised algorithms for the analysis of retroillumination images (frontal plane lens images). This includes methods of segmentation as well as several unique algorithms for calculating the centrality and integrated density of the cataract under study. We have begun to document these procedures. A second major collaborative effort with NHLBI as well as with DCRT/DSB is the measuring of blood flow velocity in arteries, and possibly through heart valves, non-invasively. Current ultrasound technology allows physicians to view flow approximately, but not quantitatively. Present systems provide a color display of flow which is actually a simulation of flow velocity. The goal for this project is to use an echo Dippler color mapping system and image the same artery or organ from multiple locations and orientations, then to sue 3D calculations and reconstruct a true flow profile. The method being developed should allow not only better calculation of velocity profiles, flow volume and resistances, but also estimations of pressures across valve orifices and stenotic arteries and for other purposes. The correction of flow velocity values on a two- dimensional images was successfully implemented this year. However, the alignment of data into a 3D and 4D volume has encountered difficulties with the accuracy of the bird 3 space tracker, which may not have the resolution to track objects that are smaller in size then the sensing crystals.