No Annotate Again (NAA): Realistic Image and Annotation Synthesis for Multi-Contrast MRI through Diffusion without Paired Data

Motivation: The variability of multi contrast MRI presents challenges when developing neural networks, as annotations are performed separately for each contrast despite their visual similarity. Goal(s): We aim to achieve scalable and automated annotation solutions for diverse MRI contrasts. Approach: We propose No Annotate Again (NAA), a novel approach that synthesizes realistic images for a new contrast using given anatomical masks, without requiring paired images or manual annotations, by designing a unique diffusionb-based framework. Results: Tested on cardiac MRI cine images and T1 maps, NAA generated realistic T1 maps, which largely improved the segmentation downstream task performance. Impact: NAA enables scalable, annotation-free neural network developments for medical image analysis. This approach reduces dependency on annotated datasets and can benefit a wide range of applications.