Journal Archive
Current Issue
ABOUT THIS IMAGE
CPDM: A Conditional Probabilistic Diffusion Model for Synthesizing Breast Cancer MRI from Integrated Multi-Omic Data Profiles
The Conditional Probabilistic Diffusion Model (CPDM) generates synthetic MRIs from multi-omic data through two pivotal phases: forward and backward diffusion. In forward diffusion, noise is progressively added to MRI images, ultimately degrading them to pure noise. In backward diffusion, the model iteratively refines the noisy images by removing the noise predicted by a cross-attention enhanced UNet module, guided by multi-omic features processed through Bayesian Tensor Factorization (BTF), to precisely reconstruct new MRIs. Chen et al
Image Credit: Lianghong Chen, Western University, 2024. Published under the Creative Commons Attribution License (CC BY 4.0).