Digital Methods for Change Detection in Medical Images

Abstract

The detection and visualisation of the changes between two images is the basis or the goal of several imaging techniques. Thus, digitized subtraction angiography consists in visualizing the differences between two images obtained without and with iodine contrast, intravenously injected. In Nuclear Medicine, the comparison of two scintigraphic images of the same organ explored under varying conditions (images acquired at different times, with different tracers, after various physiological or pharmacological interventions) is a routine problem. The visual comparison of the images is often a difficult task for the following reasons: The differences can be too low to be visually identified. In certain types of images, there are normal statistical fluctuations which can mask or simulate a difference. The gray level intensities can be different in the images which must be first normalized. Therefore, it seems reasonable to use the capacities of a computer and the knowledge about the count fluctuations to perform an automated comparison of the images. Better performances than those given by the visual inspection can be expected from such a procedure. To form this comparison, it is necessary to first register the images (alignment, normalization, magnification,…) (1,2,3,4,5) and second detect the changes by analyzing the images point by point (6,7).