My research centres on making neuroimaging more rigorous and reproducible. This spans three overlapping areas: community standards for data organisation and sharing, statistical methods for M/EEG and fMRI, and the tools and infrastructure that make open science tractable at scale.
Research Themes
Open Neuroimaging Standards
Much of the field’s reproducibility problem is upstream of analysis — in how data are collected, described, and shared. I co-lead or contribute to several community efforts that define how this should work: BIDS extensions for EEG, Genetics, and PET; the OHBM COBIDAS guidelines for M/EEG and MRI; and the Open Brain Consent templates that allow participants to authorise open sharing under GDPR. The common thread is building community consensus around practices that make individual studies reusable.
Related: COBIDAS MEEG · BIDS · Open Brain Consent
Statistical Methods for Neuroimaging
Standard parametric methods applied mass-univariately at the sensor or voxel level carry assumptions that are routinely violated in neuroimaging data — non-normality, outliers, mis-specified models, and inflated false positives from multiple comparisons. I develop and maintain tools for robust estimation (trimmed means, robust correlations, Winsorised statistics), hierarchical GLM for EEG/MEG (LIMO MEEG), adaptive thresholding for single-subject fMRI maps, and methods for single-trial ERP analysis.
Related: LIMO MEEG · SPMup · Robust Statistical Toolbox
Brain Structure, Function & Cognition
Empirically, my work spans face and object processing (ERP latencies and magnitudes), fMRI task design and signal quality, vascular parcellation of the cortex, and — more recently — transdiagnostic psychopathology and pain–reward interactions. A consistent methodological concern is whether the numbers we report are well-defined and interpretable: % signal change, effect sizes, confidence intervals, and visualisations that convey uncertainty honestly.
Selected Publications
Reproducibility & Best Practices
- Issues and recommendations from the OHBM COBIDAS MEEG committee for reproducible EEG and MEG research Nature Neuroscience
- Open and reproducible neuroimaging: from study inception to publication NeuroImage, 119623
- Improving functional magnetic resonance imaging reproducibility GigaScience, 4, 15
- Data visualization for inference in tomographic brain imaging European Journal of Neuroscience, 51, 695–705
- Improving standards in brain-behavior correlation analyses Frontiers in Human Neuroscience, 6, 119
- Can We Standardize Clinical Functional Neuroimaging Procedures? Frontiers in Neurology, 8, 1153
- Brainhack: developing a culture of open, inclusive, community-driven neuroscience Neuron, 109, 1769–1775
- Visual object categorization in the brain: what can we really learn from ERP peaks? Frontiers in Human Neuroscience, 5, 156
- Quantifying the Time Course of Visual Object Processing Using ERPs: It's Time to Up the Game Frontiers in Psychology, 2, 107
- Single-trial analyses: why bother? Frontiers in Psychology, 2, 322
Open Data & Standards
- EEG-BIDS, an extension to the brain imaging data structure for electroencephalography Scientific Data, 6, 103
- PET-BIDS, an extension to the brain imaging data structure for positron emission tomography Scientific Data, 9, 65
- The genetics-BIDS extension: Easing the search for genetic data associated with human brain imaging GigaScience, 9(10), giaa104
- The Open Brain Consent: Informing research participants and obtaining consent to share brain imaging data Human Brain Mapping
- On the Long-term Archiving of Research Data Neuroinformatics, 21, 243–246
- Improving data availability for brain image biobanking in healthy subjects: practice-based suggestions from an international multidisciplinary working group NeuroImage, 153, 399–409
- Longitudinal multi-centre brain imaging studies: guidelines and practical tips for accurate and reproducible imaging endpoints and data sharing Trials, 20, 1
- #EEGManyLabs: Investigating the Replicability of Influential EEG Experiments Cerebral Cortex
Methods & Applications
- Mixture model for single-subject fMRI thresholding Frontiers in Human Neuroscience
- BOLD signal decomposition: correcting HRF parameter estimates and computing percentage signal change Frontiers in Neuroscience
Preprints
- Poldrack, R. et al. (2023). The Past, Present, and Future of the Brain Imaging Data Structure (BIDS). arXiv.
- Randau, M. et al. (2023). Transdiagnostic psychopathology in the light of robust single-trial event-related potentials. Authorea.
- Hoskin, R., Pernet, C. & Talmi, D. (2023). Interactions between the representations of pain and reward suggest dynamic shifts in reference point. bioRxiv.