Disentangled Diffusion Autoencoder for Harmonization of Multi-site Neuroimaging Data

TL;DR

This paper introduces a novel diffusion autoencoder model for harmonizing multi-site neuroimaging data, effectively removing site effects while preserving biological variability, and demonstrating superior image quality over previous methods.

Contribution

The paper presents the first diffusion-based model for neuroimaging data harmonization, improving image quality and site effect removal compared to existing techniques.

Findings

Outperforms previous methods in generating high-resolution, harmonized MR images.

Successfully preserves biological variability during site adjustment.

Demonstrates effectiveness across data from 7 different sites.

Abstract

Combining neuroimaging datasets from multiple sites and scanners can help increase statistical power and thus provide greater insight into subtle neuroanatomical effects. However, site-specific effects pose a challenge by potentially obscuring the biological signal and introducing unwanted variance. Existing harmonization techniques, which use statistical models to remove such effects, have been shown to incompletely remove site effects while also failing to preserve biological variability. More recently, generative models using GANs or autoencoder-based approaches, have been proposed for site adjustment. However, such methods are known for instability during training or blurry image generation. In recent years, diffusion models have become increasingly popular for their ability to generate high-quality synthetic images. In this work, we introduce the disentangled diffusion autoencoder (DDAE), a novel diffusion model designed for controlling specific aspects of an image. We apply the DDAE to the task of harmonizing MR images by generating high-quality site-adjusted images that preserve biological variability. We use data from 7 different sites and demonstrate the DDAE's superiority in generating high-resolution, harmonized 2D MR images over previous approaches. As far as we are aware, this work marks the first diffusion-based model for site adjustment of neuroimaging data.