This repository provides the datasets and codes associated with the following research article:
Pavithra Elumalai#, Yasharth Yadav#, Nitin Williams*, Emil Saucan, Jürgen Jost, Areejit Samal*, Graph Ricci Curvatures Reveal Atypical Functional Connectivity in Autism Spectrum Disorder, Scientific Reports, 12(1):8295, 2022.
(# Equal contribution; * Corresponding authors)
- For the analysis of functional connectivity networks (FCNs) of the human brain, we spatially and temporally preprocess fMRI scans from the ABIDE-I dataset, using the MATLAB based CONN toolbox.
- ABIDE-I is a shared initiative by 17 individual imaging sites and compiles studies across the globe that provide raw structural and functional magnetic resonance images (MRI) of individuals with autism spectrum disorders (ASD) and age matched typically developing controls (TD). When we initiated our study, there were a total of 1112 subjects in ABIDE-I (539 ASD and 573 HC) out of which, we used 820 subjects (395 ASD and 425 HC) for our study.
- The codes that were used in our study and analysis are available in the CODES folder.
A visual guide to functional MRI preprocessing using CONN toolbox is available here:
In our study, we evaluate the utility of two discrete Ricci curvatures, Forman-Ricci curvature and Ollivier-Ricci curvature that have been recently ported to the domain of complex networks, as indicators of abnormal topological organization in resting state fMRI based functional connectivity networks (FCNs) corresponding to individuals with autism spectrum disorder (ASD).
For this investigation, we analyzed spatially and temporally preprocessed rs-fMRI images of 395 individuals with autism spectrum disorder (ASD) and 425 typically developing controls (TD), using the CONN toolbox. Subsequently, 200 regions of interest (ROIs) or nodes were defined in the brain using the Schafer atlas, and a 200 × 200 functional connectivity matrix was generated for each subject by computing the Pearson correlation coefficient between the time series of all pairs of nodes. The schematic overview of our preprocessing pipeline can be found below.
The 200 × 200 FC matrix of every subject, that was obtained at the end of the preprocessing pipeline is available in the edge list format (assuming the FC matrices as the adjacency matrices of complete graphs) in FCM directory. The files contain three columns - source, target, and correlation coefficient as edge attributes.
From the FC matrix of a given subject, we extracted the FCNs by combining two network filtering approaches.
- maximum spanning tree (MST) using Kruskal's algorithm, and
- sparsity-based thresholding
We constructed the FCNs over a wide range of edge densities between 0.02 or 2% edges and 0.5 or 50% edges, with an increment of 0.01 or 1% edges.
The FCNs used in our study are available in the edge list format in FCN directory and are further organized by edge densities. The edgelist for each FCN contains three columns - source, target, and correlation coefficient as edge attributes. Please note that all the FCNs used in our study were binarized before further analysis.
- demographic_table.tsv
Table containing the demographic information such as imaging site, unique subject identifier, age, gender, and cohort (TD/ASD) of all the 820 subjects included in our study.
- Schaefer200.txt
A list of all the 200 regions of interest (nodes) in the FCNs. Please note that these regions are encoded as integers from 0 to 199 (in order) in the FCNs provided.
An Excel workbook containing all the Supplementary Tables S1-S7 for our paper can be downloaded from Supplementary_Tables directory. A brief description of each Supplementary Table is provided as a README file inside the directory.
In case you use the codes herein, please cite the following research article:
Pavithra Elumalai#, Yasharth Yadav#, Nitin Williams*, Emil Saucan, Jürgen Jost, Areejit Samal*, Graph Ricci Curvatures Reveal Atypical Functional Connectivity in Autism Spectrum Disorder, Scientific Reports, 12(1):8295, 2022.