(1/3) SynthSeg 2.0 is finally out on PNASNews and publicly available on FreeSurfer (surfer.nmr.mgh.harvard.edu/fswiki/SynthSeg). It segments clinical brain MRI of any orientation, resolution and pulse sequence. The output is always at 1x1x1mm resolution, no matter what the input resolution is.

New TMLR paper (w/ Francisca Vasconcelos, Bobby, and Yee Whye Teh) on uncertainty quantification for low-dose CT reconstruction with implicit neural representations:

Geometric Deep Learning 🤝 Diffusion MRI 📢📢Our new midl.conference (oral) paper builds equivariant networks for Diffusion MRI where every voxel is a spherical signal while respecting: - spatial translations and rotations - per-voxel rotations arxiv.org/abs/2304.06103 (1/5)

Steffen Czolbe is presenting at #CVPR2023 today (TUE-PM-200 in person) Neuralizer: General Neuroimage Analysis without Re-Training code: github.com/SteffenCzolbe/… video: youtu.be/_wgCESDwjZI FreeSurfer MGH Martinos Center MIT Clinical and Applied Machine Learning MIT CSAIL MIT EECS #CVPR2024

Excited for midl.conference this week! I'm presenting two projects: Data Consistent Deep Rigid MRI Motion Correction: Tuesday Neuroimaging Oral (9-10:15 AM) + Poster T03 Joint Frequency and Image Space Learning for MRI Reconstruction and Analysis: Wednesday Poster W45

There's now a plethora of free AI tools online that can sharpen blurry photos 📸 but what about 3D images? Nalini Singh spoke w/ MIT News about the beauty of combining deep learning + physics when it comes to motion correction in MRI scans. news.mit.edu/2023/mit-resea…

How far can procedurally-generated training data take us? (pretty far!) AnyStar trains a *single* model on synthetic data to segment 3D blob-like instances in any unseen bioimaging/radiology dataset without retraining or interaction! neeldey.com/any-star/ (#WACV2024 🥳) 1/4

MedNeurIPS returns for 2023! sites.google.com/view/med-neuri… Consider joining us for a day at the intersection of Machine Learning and Medical Imaging, again in the NeurIPS workshops. Submissions due Sept. 29.

Efrat Shimron Kerstin Hammernik Daniel Rueckert BMRR Veronika Spieker Hannah Eichhorn Very happy to report that this comprehensive review on deep learning for retrospective motion correction in MRI by Veronika Spieker and Hannah Eichhorn IML_Helmholtz Helmholtz Munich | @HelmholtzMunich has been accepted by IEEE-TMI ! For early access please visit: ieeexplore.ieee.org/document/10285…

I’m recruiting PhD students @Duke for fall 2024! Consider applying if you’re interested in reimagining healthcare by developing novel ML/NLP methodology. I can advise students through the CS dept and the Biostats & Bioinformatics dept. Info here: gradschool.duke.edu/admissions/app…

(1/6) Our deep dive into single cell foundation models, scBERT and scGPT - now on bioRxiv! biorxiv.org/content/10.110… w/ Nalini Singh Alejandro Buendia David Sontag Gaddy Getz

Some updates: I defended my thesis and started a postdoc at Berkeley with Laura Waller. I'll spend the next few years learning about a wide range of imaging techniques and algorithms for inverse problems -- please reach out if you'd like to talk!

Sharing a link to my #MLCB2023 oral presentation, since there were some audio issues on the livestream. youtube.com/watch?v=LI7e2c…. Thanks to the organizers for a fantastic conference!

New preprint on Fully Convolutional Slice-to-Volume Reconstruction for Single-Stack MRI arxiv.org/abs/2312.03102 Following recent single-view depth estimation methods, it produces SOTA reconstructions even with extreme inter-slice motion, eg fetal MR Work spearheaded by Sean Young

Happy to share our latest work on shape-aware segmentation has been accepted to Machine Learning for Biomedical Imaging! In this work, we tackle the challenging problem of segmenting placental volumes in whole-uterus BOLD MRI Time Series.