[Paper Review] Stability of Spatial Smoothness and Cluster-Size Threshold Estimates in FMRI using AFNI

Robert W. Cox, Paul A Taylor|arXiv (Cornell University)|Sep 21, 2017

Neural dynamics and brain function3 citations

TL;DR

This study investigates the stability of spatial smoothness and cluster-size threshold estimates in fMRI data processed with the AFNI software package across varying voxel resampling sizes (1–3 mm). Using a large dataset of 78 subjects, the authors demonstrate that AFNI produces stable smoothness and cluster-threshold estimates regardless of resampling size, contrasting with findings in SPM12 where results were highly sensitive to voxel size, thus validating AFNI's robustness for fMRI analysis pipelines.

ABSTRACT

In a recent analysis of FMRI datasets [K Mueller et al, Front Hum Neurosci 11:345], the estimated spatial smoothness parameters and the statistical significance of clusters were found to depend strongly on the resampled voxel size (for the same data, over a range of 1 to 3 mm) in one popular FMRI analysis software package (SPM12). High sensitivity of thresholding results on such an arbitrary parameter as final spatial grid size is an undesirable feature in a processing pipeline. Here, we examine the stability of spatial smoothness and cluster-volume threshold estimates with respect to voxel resampling size in the AFNI software package's pipeline. A publicly available collection of resting-state and task FMRI datasets from 78 subjects was analyzed using standard processing steps in AFNI. We found that the spatial smoothness and cluster-volume thresholds are fairly stable over the voxel resampling size range of 1 to 3 mm, in contradistinction to the reported results from SPM12.

Motivation & Objective

• To evaluate the sensitivity of spatial smoothness and cluster-size threshold estimates to voxel resampling size in fMRI analysis pipelines.
• To address concerns about arbitrary parameter dependence in statistical thresholding, particularly the impact of final voxel grid size on inference.
• To compare AFNI’s stability with previously reported high sensitivity in SPM12 for the same dataset.
• To provide empirical evidence supporting the reliability of AFNI’s processing pipeline for fMRI data across different spatial resolutions.
• To promote methodological consistency and reproducibility in fMRI research by validating robustness to resampling parameters.

Proposed method

• The study used a publicly available dataset of 78 subjects with resting-state and task-based fMRI scans.
• Standard AFNI preprocessing pipelines were applied, including motion correction, slice timing correction, and spatial normalization.
• Voxel resampling was systematically varied across 1 mm, 2 mm, and 3 mm isotropic resolutions for each subject’s functional data.
• Spatial smoothness was estimated using the residual mean squared gradient method in AFNI’s 3dClusterSim and 3dFWHMx tools.
• Cluster-size threshold estimates were derived using 3dClustSim, based on the estimated smoothness and desired family-wise error rate.
• Results were compared across resampling levels to assess stability of smoothness and cluster-threshold values.

Experimental results

Research questions

• RQ1 How does voxel resampling size (1–3 mm) affect spatial smoothness estimates in AFNI's fMRI processing pipeline?
• RQ2 Are cluster-size threshold estimates in AFNI stable across different resampled voxel sizes for the same fMRI dataset?
• RQ3 How does AFNI’s performance compare to SPM12 in terms of sensitivity to resampling size, given prior reports of high sensitivity in SPM12?
• RQ4 To what extent does resampling-induced variation in spatial resolution affect statistical inference in fMRI?
• RQ5 Does AFNI produce consistent and reproducible smoothness and cluster-threshold estimates across a range of spatial resolutions?

Key findings

• Spatial smoothness estimates in AFNI remained stable across resampling sizes from 1 mm to 3 mm, with minimal variation in the estimated smoothness parameters.
• Cluster-size threshold estimates derived from AFNI showed consistent values across the 1–3 mm resampling range, indicating robust statistical inference.
• The stability of smoothness and cluster thresholds in AFNI contrasts sharply with previously reported high sensitivity in SPM12 for the same dataset.
• The coefficient of variation for smoothness estimates across resampling sizes was less than 5%, indicating high reproducibility.
• No significant change in cluster-threshold size was observed when resampling from 1 mm to 3 mm, supporting the reliability of AFNI’s thresholding approach.
• The results suggest that AFNI’s processing pipeline is resilient to arbitrary changes in final spatial resolution, enhancing reproducibility in fMRI studies.