Fuzzy logic for the segmentation of myocardial tomoscintigraphic images

摘要

In scintigraphy, segmentation is usually based on the selection of a cut-off level applied to the original image which shows a distribution of radioactivity. We propose, for myocardial scintigraphy, a new approach in which segmentation is applied to a functional image representing the "probability of belonging to the myocardium". This image, generated by a fuzzy logic algorithm using the morphological skeleton as an input data, displays the myocardium with a higher contrast and sharper limits than on the original image and is therefore more easily segmented. Moreover any defect becomes filled. This principle was used to analyse with an extremely high reproducibility the radiotracer distribution on tomographic views of the myocardium that were previously reconstructed, reoriented and finally sectorized.