Mapping distinct timescales of functional interactions among brain networks
Authors: Mali Sundaresan, Arshed Nabeel, Devarajan Sridharan
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | First, we show, with simulated f MRI data, that instantaneous and lag-based GC identify distinct timescales and complementary patterns of functional connectivity. Next, we analyze f MRI scans from 500 subjects and show that a linear classifier trained on either instantaneous or lag-based GC connectivity reliably distinguishes task versus rest brain states, with 80-85% cross-validation accuracy. |
| Researcher Affiliation | Academia | 1Center for Neuroscience, Indian Institute of Science, Bangalore 2Department of Computer Science and Automation, Indian Institute of Science, Bangalore |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper mentions using 'MVGC toolbox' [1][19] but does not provide specific access to open-source code for the authors' own methodology. |
| Open Datasets | Yes | We analyzed minimally preprocessed brain scans of 500 subjects, drawn from the Human Connectome Project (HCP) database [12]. |
| Dataset Splits | Yes | We used Matlab s fitclinear function, optimizing hyperparameters using a 5-fold approach: by estimating hyperparameters with five sets of 100 subjects in turn, and measuring classification accuracies with the remaining 400 subjects; the only exception was for the classification analysis with averaging GC matrices (Fig. 3B) for which classification was run with default hyperparameters (regularization strength = 1/(cardinality of training-set), ridge penalty). The number of features for i GC-based classification was 91 (upper triangular portion of the symmetric 14 14 i GC matrix) and for d GC-based classification was 182 (all entries of the 14 14 d GC matrix, barring self-connections on the main diagonal). Based on these functional connectivity features, we asked if we could reliably predict the task condition (e.g. language versus resting). Classification performance was tested with leave-one-out and k-fold crossvalidation. |
| Hardware Specification | No | The paper does not specify any particular hardware (CPU, GPU models, or specific machine types) used for running the experiments. |
| Software Dependencies | Yes | Network time series were computed by averaging time series across all voxels in a given network using Matlab and SPM8. These multivariate network time series were then fit with an MVAR model (Supplementary Material Section S1). Model order was determined with the Akaike Information Criterion for each subject, was typically 1, and did not change with further downsampling of the data (see next section). The MVAR model fit was then used to estimate both an instantaneous connectivity matrix using i GC (Fx y) and a lag-based connectivity matrix using d GC (Fx y). |
| Experiment Setup | Yes | Model order was determined with the Akaike Information Criterion for each subject, was typically 1... We used Matlab s fitclinear function, optimizing hyperparameters using a 5-fold approach: by estimating hyperparameters with five sets of 100 subjects in turn, and measuring classification accuracies with the remaining 400 subjects; the only exception was for the classification analysis with averaging GC matrices (Fig. 3B) for which classification was run with default hyperparameters (regularization strength = 1/(cardinality of training-set), ridge penalty). |