A major problem in medical imaging is the inability to adequately separate interfering anatomical structure or physiological processes. This inability to isolate features of interest is particularly true in nuclear medicine imaging, in which the uptake or passage of radiopharmaceuticals in overlapping organs may interfere with the observation of the feature of interest. An image processing technique has been developed which has the ability to isolate a feature of interest from interfering features when applied to a sequence of images from the same anatomical site. The objective of this project is to investigate and evaluate the application of this technique (Eigenimage Filtering) to nuclear medicine temporal image sequences for the purpose of isolating anatomical structures and/or specific physiological processes of interest. The technique generates a composite image in which a desired feature is segmented and one or more interfering features have been removed. This composite image can be used to obtain anatomical structure information or as a functional image in the sense that a specific physiological process was taken as the desired feature. In addition, this composite image can be used with the original image sequence to regenerate a temporal image sequence in which a physiological process is isolated from those that interfere with its observation. This regenerated sequence technique will be used in the quantitative analysis of physiological information from first pass studies of the cardiac system and transplanted kidneys. In the first pass cardiac studies, the left ventricular ejection fraction determined from the regenerated sequence will be compared to that obtained from equilibrium gated blood pool scintigraphy and from cardiac catheterization with contrast ventriculography. In the transplanted kidneys studies, the renal perfusion index will be determined and its accuracy in determining the etiology of renal failure will be compared to percutaneous needle biopsy results and clinical observation (rejection). The ability to segment features of interest from those that interfere with their observation can significantly improve the diagnostic information available to clinicians.
