From Theory to Treatment

Advancing Diffusion Models for Medical Diagnosis

Accelerate-@CambridgeC2D3 Funding Awarded Project



Imaging is a key tool in medical diagnosis but images and be ‘noisy’, containing multiple sources of information and are sometimes incomplete. The multidisciplinary project team will pioneer novel adaptive sampling techniques to streamline the training process of diffusion models used in image analysis. This innovation will significantly reduce computation resources, rendering diffusion models more practical for real-world applications, including medical imaging diagnosis.

This project was supported with funding from the Cambridge Centre for Data-Driven Discovery and Accelerate Programme for Scientific Discovery, made possible by a Donation from Schmidt Futures.



People


Research Leader


Angelica I Aviles-Rivero
University of Cambridge


Research Assistant


Yanqi Cheng
University of Cambridge


Collaborators

Alphabetic Order

Jiahao Huang
Imperial College London
Carola-Bibiane Schonlieb
University of Cambridge
Shujun Wang
The Hong Kong Polytechnic University
Guang Yang
Imperial College London
Lequan Yu
The University of Hong Kong
Chun-Wun Cheng
The University of Cambridge






Publication


The Missing U for Efficient Diffusion Models
S CalvoOrdonez, C-W Cheng, J Huang, L Zhang, G Yang, CB Schonlieb, AI Aviles-Rivero
Transactions on Machine Learning Research (TMLR) || Arxiv-Link

Single-Shot Plug-and-Play Methods for Inverse Problems
Y Cheng, L Zhang*, Z Shen*, S Wang, L Yu, RH Chan, C-B Schonlieb and AI Aviles-Rivero
Preprint Version — Arxiv-Link

Beyond U: Making Diffusion Models Faster & Lighter
S CalvoOrdonez, J Huang, L Zhang, G Yang, CB Schonlieb, AI Aviles-Rivero
NeurIPS 2023 W/ on Diffusion Models
Arxiv-Link || Conference || Code


Prior Work


CDiffMR: Can We Replace the Gaussian Noise with K-Space Undersampling for Fast MRI?
J Huang,  AI Aviles-Rivero,  C-B Schönlieb and G Yang
MICCAI 2023
Arxiv-Link || Conference || Code

DiffMIC: Dual-Guidance Diffusion Network for Medical Image Classification. Y Yang, H Fu,  AI Aviles-Rivero,  C-B Schönlieb and L Zhu
MICCAI 2023 (early accept)
Arxiv-Link || Conference || Code

TFPnP: Tuning-free Plug-and-Play Proximal Algorithm with Applications to Inverse Imaging Problems.
K. Wei, AI Aviles-Rivero, J Liang, Y. Fu, H Huang and C-B Schönlieb
Journal of Machine Learning Research (JMLR)  Arxiv-Link || Journal Link || Code Repository
🏆Extended Version of our ICML 2020 Outstanding Paper Award




Activities


Research Retreat with collaborators -- Cambridge January 2024