Study of Various Neural Networks to Improve the Defuzzification of Fuzzy Clustering Algorithms for ROIs Detection in Lung CTs

摘要

The detection of pulmonary nodules in CT images has been extensively researched because it is a highly complicated and socially interesting matter. The classical approach consists in the development of a computer-aided diagnosis (CAD) system that indicates, in phases, the presence or absence of nodules. A common phase of these systems is the detection of regions of interest (ROIs), that may correspond to nodules, in order to reduce the searching space. This paper evaluates the us& of various neural networks for the defuzzification of the output of fuzzy clustering algorithms, in order to improve the detection of true positives and the reduction of false positives. Also, they are compared to the results from a support vector machine (SVM).