Thomas E. Nichols

This paper develops a Bayesian spline method to model the spatio-temporal distribution of white matter hyperintensities with respect to age and validates it on simulated and real datasets.

The paper revisits cluster inference methods in fMRI, finding that data cleaning and two-sided tests reduce false positive rates, and estimates 10% of fMRI studies use problematic methods.

The paper critiques Narum's implementation of the Benjamini-Yekutieli procedure for family-wise error correction, showing it fails to adequately control for false positives.

The paper develops a method for efficient genome-wide association analysis of high-dimensional imaging phenotypes using mixed models and variance component estimation.

This paper introduces dynamic graphical models for directed functional connectivity and evaluates their performance using simulated and human fMRI data.

This paper introduces a variance decomposition framework for rs-fMRI data that redefines DVARS and provides statistical methods to assess scan quality.

The paper shows that joint analysis of cortical area and thickness is more sensitive than volume for studying brain anatomy, and proposes a permutation-based method to replace volume analysis.

This paper outlines best practices for open and reproducible neuroimaging research using MRI, providing guidelines for data analysis, sharing, and reporting.

The paper evaluates 15 candidate SNPs associated with psychiatric disorders using 6,165 diffusion MRI scans, finding no significant associations after correction for multiple testing.

The paper provides practical recommendations for improving data availability and sharing in brain image biobanking to enhance research and clinical applications.

This paper develops a Bayesian mixed model to address the stimulus-as-fixed-effect fallacy in fMRI, showing it reduces false positives and improves statistical inference.

This paper introduces NIDM-Results, a standardized format for sharing neuroimaging statistical results, enabling reproducibility and meta-analysis.

The paper introduces BIDS, a standard for organizing MRI datasets to improve data sharing and processing.

This paper introduces a method to calculate power and sample size for fMRI studies using prevalence of active peaks and effect size, validated with simulations and real data.

This paper evaluates eight meta-analytic approaches for image-based fMRI meta-analysis using simulated and real data, finding that combining contrast estimates with random-effects models or non-parametric statistics provides the best accuracy.

NeuroVault provides a web-based repository for storing, sharing, and decoding unthresholded brain statistical maps with a public API.