Automated registration of planar images for improved diagnosis, region-of-interest analysis, and dosimetry

Organ and target tissue pharmacokinetics for dosimetry are usually obtained from a longitudinal series of planar gamma camera images. The information is reduced to a series of individual time-activity curves, where each curve represents the mean change in radioactivity concentration over time in a user-defined region-of-interest. We present a method for automated extraction of such information which also converts the longitudinal image set into a single parametric {open_quotes}clearance-rate image{close_quotes} over the whole body. Each of the posterior images was first inverted to create a pseudo-anterior image. Each pseudo-anterior image was then registered to the original anterior image using a symmetric phase-only matched filtering (SPOMF) technique. The transformations required to register each of these image sets to each other over time were obtained, again using SPOMF, but applied to a selected set of sub-regions that exhibited a consistent pattern of radioactivity distribution over time. After the original image set has been modified in this way, the extraction of time-activity-curves for regions-of-interest becomes simpler since a single set of contours will apply to all of the images. By fitting each individual pixel value in the modified set of images, to an exponential function over time, the time-sequential images may be represented bymore » a parametric image of radioactivity clearance rate. Application of this methodology to planar images obtained for radioiodine treatment of thyroid cancer yielded valuable diagnostic and therapeutic information. From the parametric {open_quotes}clearance-rate image,{close_quotes} the stomach was clearly identified as a region of rapid radioiodine clearance and the liver demonstrated continued uptake over the image collection period, suggesting continued production of thyroid hormone by the metastases. Individual tumor metastases showed both uptake and clearance over the 4-day image collection period.« less