Resting state networks characterization for individual subjects assessment

Abstract

Cumulative research in hemodynamic brain activity measured in resting state (RS) using functional magnetic resonance imaging (fMRI) suggests that healthy brain dynamics are distributed on large-scale spatial resting state networks (RSNs). These networks correspond to well-defined functional entities that have been associated to different low and high brain order functions. Characterization of several pathological and pharmacological conditions have been studied by measuring the changes introduced in the RSNs by these affections, resulting on powerful and descriptive biomarkers. Most of these studies have been performed using methods devised for group level analysis. Nevertheless, the use of these biomarkers in diagnostic/prognostic tasks may require single subject level assessment. In addition, some brain conditions are characterized by a high intra-subject variability, which violates the underlying assumptions of most of the group based methods. Recently, a multiple template matching (MTM) approach was proposed to automatically recognize RSNs in individuals subject’s data. This method provides valuable information to assess subjects at individual level. In this work we propose a set of changes to the original MTM that improves the RSNs recognition task and also extends the functionality of the method. The key points of this improvement are: An standardization strategy and a modification of the method’s constraints in order to add flexibility. Additionally, we also present a novel approach to quantify the degree of trustworthiness for each RSN obtained by using template matching. The main idea is to use a double validation process in the following way: First, RSNs are identified in a curated dataset which we’ll call subjects of reference. Second, we propose to use these subjects of reference along with MTM to validate how much the template’s assignations coincide. Finally, we integrate these solutions into an open source framework built on top of one of the most popular tools used by the community. Our results suggest that the method will provide complementary information for characterization of RSNs at individual level.