Improved Generation of Synthetic Imaging Data Using Feature-Aligned Diffusion

TL;DR

This paper introduces feature-aligned diffusion, a method that improves synthetic medical image generation by aligning model features with expert features, resulting in higher accuracy and diversity.

Contribution

The paper proposes a novel feature-aligned diffusion approach that enhances synthetic image quality and can be integrated with existing diffusion models.

Findings

9% improvement in generation accuracy

~0.12 increase in SSIM diversity

Easily integrated into existing diffusion pipelines

Abstract

Synthetic data generation is an important application of machine learning in the field of medical imaging. While existing approaches have successfully applied fine-tuned diffusion models for synthesizing medical images, we explore potential improvements to this pipeline through feature-aligned diffusion. Our approach aligns intermediate features of the diffusion model to the output features of an expert, and our preliminary findings show an improvement of 9% in generation accuracy and ~0.12 in SSIM diversity. Our approach is also synergistic with existing methods, and easily integrated into diffusion training pipelines for improvements. We make our code available at \url{https://github.com/lnairGT/Feature-Aligned-Diffusion}.