Insulto srities atpažinimo kompiuterinėse tomogramose algoritmų efektyvumo palyginimas

Abstract

There are developed several computer aided methods for recognition of ischemic stroke region on CT slices at this time. This paper discuss about evaluation of such recognition methods with the help of human experts. Presented evaluation is completed by using an algorithm of fuzzy sets. Performed analysis of 20 methods with modifications showed, that gray level co-occurrence matrix method and hybrid method give highest rates. In addition, the evaluations of the same images by human experts are not uniform. More complex geometrical evaluation methods must be used for archiving of the better accuracy.