Construction of embedded fMRI resting state functional connectivity networks using manifold learning

TL;DR

This paper introduces a method for constructing embedded functional connectivity networks from rsfMRI data using manifold learning algorithms, and evaluates their effectiveness in classifying schizophrenia versus healthy controls.

Contribution

It compares linear and nonlinear manifold learning algorithms for FCN construction and assesses their classification performance with machine learning.

Findings

Diffusion Maps with lagged cross-correlation outperform other methods

Embedded FCN effectively distinguish schizophrenia from controls

Manifold learning enhances functional connectivity analysis

Abstract

We construct embedded functional connectivity networks (FCN) from benchmark resting-state functional magnetic resonance imaging (rsfMRI) data acquired from patients with schizophrenia and healthy controls based on linear and nonlinear manifold learning algorithms, namely, Multidimensional Scaling (MDS), Isometric Feature Mapping (ISOMAP) and Diffusion Maps. Furthermore, based on key global graph-theoretical properties of the embedded FCN, we compare their classification potential using machine learning techniques. We also assess the performance of two metrics that are widely used for the construction of FCN from fMRI, namely the Euclidean distance and the lagged cross-correlation metric. We show that the FCN constructed with Diffusion Maps and the lagged cross-correlation metric outperform the other combinations.