Reliability test for processing of magnetic resonance images in resting state using graph theory

A large number of current studies have been conducted using graph theory as magnetic resonance image processing. Some of the methods used in graph theory can lead to important differences when it comes to getting results. Preprocessing variables, the atlases used or correlations influence the results that will later be analysed. This paper shows a reliability study on resting state imaging of control patients. Five different methods are presented to conduct the study with graph theory. By using different atlases and correlation methods, the graph measures most currently used are calculated. The results of the measures are analysed using the intraclass correlation coefficient. The reliability of the measures is evaluated in relation to the proposed methods. Measures such as degree, clustering, strength, modularity, path length and global efficiency always show a reliability of over 70%. This means that its use in other pathology evaluation studies provides greater reliability with regard to result analysis.