This project is an extension of previous work involving the creation, development and testing of image processing techniques designed to improve diagnostic performance. Current work has centered on methods relevant to the processes of radiologic image subtraction and tomosynthesis. Work continues in the area of automated manipulation to facilitate registration. Particular emphasis has been placed on 1) methods for correcting contrast differences existing between seriallly obtained radiographic image projections, 2) methods for eliminating noise produced by quantum fluctuations in radiographs, 3) methods for quantifyng the apparent size of lesions from associated diagnostic images and 4) methods for precisely manipulatng digital images in ways essential for registration preparatory for image subtraction. A new nonparametric algorithm for correcting differences in radio-graphic contrast has been developed which is simpler, faster, and more robust than its parametric counterpart described last year. Another computer program has been developed which permits quantum fluctutions intrinsic to low dose radiographic systems to be eliminated by means of rapid median filtration using available frame-buffer hardware. Improved measures of image similarity and techniques for quantifying the size of lesions from subtracted radiographs facilitate unbiased image analysis essential to quantitative radiology within the constraints of a clinically practical system. The latter makes use of a lesion perimeter detection scheme based on radially determined gradient measurements abstracted from the subtraction image. By comparing contrast exhibited by the lesion relative to its background in histogram-compensated images, correlations of better than 0.95 have been made with nodular defects of known size.
