PhD student @Parietal, Machine Learning x Neuroimaging
Paris
Joined September 2019
- Tweets19
- Following297
- Followers54
- Likes40
Thomas Bazeille retweeted
We’ve just posted a revised version of this preprint ! @tbazeill @jbpoline @BertrandThirion
Quick highlights 🧵 below
xitter.gq/emdupre_/status/133671…
Excited to share our newest work, co-1st authored with @tbazeill!
preprint: biorxiv.org/content/10.1101/…
code: github.com/neurodatascience/…
With @jbpoline @BertrandThirion, we benchmarked how functional alignment can improve inter-subject #fMRI decoding performance.
#tweeprint thread 1/
ALT Academic figure showing a qualitative visualization of the impact of different functional alignment transformations on fMRI signal structure.
Thomas Bazeille retweeted
New paper in @NeuroImage_EiC ✨ "Decoding with Confidence: Statistical Control on Decoder Maps"
sciencedirect.com/science/ar…
with Jerome-Alexis Chevalier, @_tbng, @salmonjsph & @BertrandThirion
Why statistical control on decoder maps? What kind of control? How? 1/5
⤵️
Thomas Bazeille retweeted
Really happy to share that I got 2 abstracts accepted at the @ISMRM 2021 conference for my first submissions, one of which is going to be an oral in the ML for Image Reconstruction session !
Thomas Bazeille retweeted
🎉 @mne_python 0.22 is out! 🎉
✅ label selections in the source time course viewer (pic)
✅ #sEEG & #ECoG visualization
✅ simplified ICA
✅ better volumetric source space support
✅ spatio-spectral decomposition
✅ & more!
👉 mne.tools/stable/whats_new.h…
#mne #eeg #meg #neuroimaging
Excited to share our newest work, co-1st authored with @emdupre_ !
Preprint: biorxiv.org/content/10.1101/…
code: github.com/neurodatascience/…
With @jbpoline @BertrandThirion, we benchmarked how functional alignment can improve inter-subject #fMRI decoding performance
#tweeprint thread 1/
Additionally, we find that Shared Response Model shows the strongest performance at the ROI-level.
To learn full-brain alignments, we also show that "piecewise" (i.e., non-overlapping ROIs) alignment performs better than a searchlight alignment scheme 6/
Thank you to all of the investigators who shared their data to make this possible, incl. @JosetAEtzel, @TakerkartSylva1, @studyforrest, @Parietal_INRIA, and @CNeuromod
Our code, too, builds on work from many others, incl. @scikit_learn, #nilearn, @pymvpa, and @UseBrainIAK ! 7/7
Thomas Bazeille retweeted
We're preparing to release #fMRIPrep 20.2, which will be a long-term support (LTS) release. LTS means this version of fMRIPrep will be supported for five years, to provide a stable preprocessing base for long-running studies. 1/
Thomas Bazeille retweeted
Our paper on multi-subject M-EEG source imaging has (finally) appeared at Neuroimage 🥳 OT + sparsity -> increased similarity with fMRI activation patterns.
sciencedirect.com/science/ar…
Python code: github.com/hichamjanati/neur…
with @tbazeill @CuturiMarco @BertrandThirion @agramfort
Does hyperalignment improve inter-subject decoding ? Spoiler : some methods do! We provide experimental results at #OHBM2020Posters n°1382. Happy to discuss it!
cc @emdupre_ @jbpoline @BertrandThirion @Parietal_INRIA
Thomas Bazeille retweeted
Announcing the dev nilearn days: software and science
19-22 May 2020: mark the date
Join online the #Nilearn development community and shape what's next
nilearn.github.io/dev-days-2…
Our goal is to grow a stronger technical community for statistics on brain images in Python
Thomas Bazeille retweeted
.@tbazeill @BertrandThirion did a great job showcasing it against RR, Procrustes for fMRI in their recent paper hal.archives-ouvertes.fr/hal…
And they have a scikit-style implementation, too ! github.com/Parietal-INRIA/fm…