Enigma Center for Worldwide Medicine, Imaging & Genomics

06/11/2025

Talairach Lecture - huge thanks to OHBM for putting this (very nervous) lecture online!
https://www.youtube.com/watch?v=3iWpveOI-0E

OHBM 2025 Keynote Lecture
Speaker: Paul Thompson
Title: Talairach Lecture: Worldwide Collaboration and Artificial Intelligence in the Global Quest to Map Human Brain Diseases.
Abstract: Since 2009, the ENIGMA Consortium has brought together over 2,000 brain researchers from 47 countries to coordinate the largest neuroimaging studies of over 30 brain diseases – a global effort driven by the expertise and vision of many in this very audience. This global alliance reflects the shared vision, mutual support, and collective power of the neuroimaging community, to tackle questions that no single lab can address on its own. ENIGMA began by tackling the replication crisis in brain imaging genetics, uncovering over 500 common and rare genetic and epigenetic variants that influence brain structure, function, and the speed of brain aging (Grasby et al., Science, 2020; Brouwer et al., Nature Neuroscience, 2022). These findings laid the groundwork for broader studies: today, ENIGMA’s 30 disease-focused working groups coordinate the largest MRI, DTI, and fMRI studies in psychiatry (e.g., schizophrenia, bipolar disorder, major depression, PTSD, addiction, OCD), neurology (e.g., epilepsy, Parkinson’s, ataxia, neuro-HIV, TBI, chronic pain), and several neurodevelopmental brain disorders (anorexia, conduct disorder, Tourette syndrome, neurogenetic syndromes). By applying rigorous, consensus-driven workflows to vast global datasets, ENIGMA is revealing unexpected disease subtypes, risk factors, and treatment effects, identifying emerging principles that connect and distinguish multiple brain disorders, linking to neuroscientific findings at cellular and molecular scales (Larivière et al., ENIGMA Toolbox, Nature Methods, 2021).
ENIGMA’s impact extends globally through initiatives such as ENIGMA India, Pakistan, and the ENIGMA-U education program, fostering open collaboration, training, and opportunities for all. In this talk, we will outline major findings to date and examine how AI-driven methods may reshape the landscape of brain research. With generative AI, vision-language models, and multimodal brain decoding, we are on the cusp of new possibilities – from decoding disease mechanisms to designing targeted treatments. The global quest to alleviate brain diseases is within our reach – a collective endeavour that is greatly empowered by the OHBM community, supporting us all in our shared mission to bring life-changing discoveries to all corners of the globe.

OHBM 2025 Keynote LectureSpeaker: Paul ThompsonTitle: Talairach Lecture: Worldwide Collaboration and Artificial Intelligence in the Global Quest to Map Human...

19/09/2024

From Blood Tests to Brain Scans: How is Revolutionizing Alzheimer’s Research - https://viterbischool.usc.edu/news/2024/09/from-blood-tests-to-brain-scans-how-ai-is-revolutionizing-alzheimers-research/

1000000 thanks to Omar Lewis for covering this!

18/09/2024

What an exhilarating day of mathematics at UPenn Big Data Conference focusing on multimodal, multisite data integration to discover and combine biomarkers of disease. There was a connected mathematical trajectory of ideas, starting with:
1. merging sc-RNAseq, ATAC-seq + proteomics across datasets without eliminating disease effects (Prof Nancy Zhang [1] )
2. Rong Ma's spectral harmonization methods where you align batches of high-dimensional data by embedding them and testing the alignment of their spectra (eigenvalue sets)
3. Conditional latent diffusion models (LDMs), where you generate one modality, conditioned using another modality (a text prompt), vectors, or a source image ( https://pubmed.ncbi.nlm.nih.gov/38370616/ )
4. Heng Huang's method [3] to train a single-modality deep model using a "teacher" that is a multimodal model; here you can use genomic data to train an AI model that will only be using images!
5. Hongtu Zhu's () multivariate imaging genomics algorithms for voxel-wise GWAS and GWAS of genetically correlated traits [4]
6. 3D SAM (Segment Anything Model) as a foundation model for segmenting all 3D imaging data and surgical navigation videos [5] (Mathias Unberath).
A big thank you to Li Shen for the invitation !
[1] https://statistics.wharton.upenn.edu/profile/nzh/
[2] https://web.cvent.com/event/47f55e50-9b98-493a-bc42-58123641c5e5/websitePage:c6af7a53-a55f-4161-a0c3-e1606a28b43d
[3] https://arxiv.org/abs/2305.16222
[4] https://arxiv.org/abs/2210.09217
[5] https://arxiv.org/abs/2402.17972

16/09/2024

Standing-room only at today's jam-packed "AI in Neurology" Presidential Plenary [1,2] at the American Neurological Association showcasing AI for image-based diagnosis, remote patient monitoring, patient video analysis + more. Vast queue of questioners emphasized need to create AI methods to respond adaptively to a continuous intermittent multimodal data stream (e.g., reinforcement learning in oncology [3]):
[1] Newswire: https://newswise.com/articles/artificial-intelligence-poised-to-transform-neurological-care-from-diagnosis-to-treatment-even-prevention
[2] Organised by Paralympian Prof Cassie Mitchell who represented the USA in Paris last week [4] and still presented s-SuStaIn for staging diseases based on clustering and staging of high-dimensional biomarkers https://proceedings.mlr.press/v248/tandon24a.html
[3]https://ncbi.nlm.nih.gov/pmc/articles/PMC8472712/pdf/cancers-13-04624.pdf
[4] https://coe.gatech.edu/news/2024/08/cassie-mitchell-pursues-4th-paralympic-medal-her-4th-straight-games

22/06/2024

COME JOIN US TOMORROW @ "Deep Learning in Neuroimaging"

Sunday June 23 9am-5:30pm COEX Center Room E5 Seoul, Korea

Learn Generative AI, other hot-topics

https://ww6.aievolution.com/hbm2401/index.cfm?do=ev.viewEv&ev=1559
Register: https://humanbrainmapping.org/i4a/pages/index.cfm?pageid=4249
Quick Intro: https://youtube.com/watch?v=fOORfzGjCTA&t=1615s

All-day action-packed educational course: feat. Anqi Qiu Tom Nichols Chao-Gan Yan Sergey Plis, Jing Sui !

15/06/2024

Generative AI without Tears
https://www.youtube.com/watch?v=uh1Jw3eH8FA
1:02 What is Generative AI?
1:30 What is a variational autoencoder (VAE)?
2:15 Modeling variations in handwritten digits
4:06 Training a VAE
5:18 Realistic Inpainting with a VAE
6:10 A Face Generator
6:30 What is the Perceptual Loss?
8:23 Can you Generate Brains?
10:43 A person eating noodles
12:00 Langevin diffusion
13:35 Score functions and reversing noise in an image
14:03 The Fokker-Planck analog of the Langevin diffusion
14:30 Contextual Guidance
16:00 Turning the stochastic term off: DDIM
19:30 Cycle-GANs: Another conditional generative model of images
20:10 Neural Style Transfer

15/06/2024

🔥 NEW VIDEO TUTORIAL 🔥 (ha, yes I honestly wrote a new lecture !): ...A Guided Tour of AI in Neuroimaging.
https://www.youtube.com/watch?v=fOORfzGjCTA
Topics!
6:30 Applications of Deep Learning
12:08 Convolutional Neural Networks (CNNs)
13:00 Backpropagation
14:53 Occlusion Sensitivity Analysis
16:35 Grad-CAM
20:07 CNN-LSTM
21:30 Vision Transformers
22:55 Pre-Training on ImageNet
23:46 The ImageNet Challenge
26:45 Generative AI!
27:47 Variational Autoencoders (VAEs)
33:11 Perceptual Loss, Face Generation, and Synthetic MRI Generation
38:35 Denoising Diffusion Probabilistic Models (DDPMs)
44:35 AI and the Connectome
46:28 Neural Style Transfer
56:10 Generative Adversarial Networks (GANs)
I hope you enjoy it !! 😊

26/04/2024

Finally made it to NIMHANS, Bangalore, to work on our India ENIGMA study of factors that affect brain aging and mental health in India. Congrats to all the new PhD students (Mekha, Aarthi, Gauthami) for presenting their wonderful PhD projects applying novel AI methods to the India ENIGMA MRI scans ! https://reporter.nih.gov/search/x2BnLSMTFU-UkxKpOBB4vw/project-details/10187479

23/04/2024

If you’re at come see our talk at 9am !

20/04/2024

My daughter said to definitely not bring this on the plane, but I did anyway. :) It's a pollution meter recommended by Dr Brittany Taylor (expert on pollution and the brain). We'll test it in our India ENIGMA Initiative. So far, higher pollution correlates with better sunset photos :) PM2.5 was 38 μg/m^3 in L.A. on Thursday and 81 μg/m^3 rising to 118 μg/m^3 right now in Abu Dhabi (Saturday night)...

30/03/2024

Amazing experience today to catch up for 3.5 hours with Prof Guojun Liao who I last met 27 years ago in 1997 as a PhD student at UCLA. His variational methods for analysing 3D surfaces and diffeomorphic flows helped me get into neuroimaging after being a bit lost about how to get a job using mathematics [1]. Today, he taught us fractional calculus and why we should compute Curl for diffusion tensor fields ! And he remembered our work making surface meshes that make a surface-based signal stationary and isotropic (well L2-close). A very humble person with very creative ideas !
[1]https://pages.stat.wisc.edu/~mchung/teaching/MIA/reading/SBM.thompson.2004.pdf

09/02/2024

This was a fun project to work on, and we are hoping to try it out in ENIGMA: Can you use Artificial Intelligence to "reverse" the disease process in your brain MRI + highlight your brain’s abnormalities ? Nikhil Dhinagar in our lab explains how you can encode high-dimensional healthy brain variation using a denoising diffusion probabilistic model (DDPM) with context vector guidance in the latent space to highlight disease effects in your MRI scan. Here's the paper on the new technique:https://www.biorxiv.org/content/10.1101/2024.02.05.578983v1.full.pdf