GM-LDM: Latent Diffusion Model for Brain Biomarker Identification through Functional Data-Driven Gray Matter Synthesis

TL;DR

GM-LDM introduces a novel latent diffusion model framework utilizing a 3D autoencoder and Vision Transformer to improve MRI generation, enabling personalized brain biomarker identification and disease analysis.

Contribution

The paper presents GM-LDM, a new latent diffusion model framework that integrates a 3D autoencoder and ViT-based denoising network for enhanced MRI synthesis and brain biomarker discovery.

Findings

Achieves statistical consistency in MRI generation.

Enables personalized brain imaging with functional data.

Facilitates biomarker identification for brain diseases.

Abstract

Generative models based on deep learning have shown significant potential in medical imaging, particularly for modality transformation and multimodal fusion in MRI-based brain imaging. This study introduces GM-LDM, a novel framework that leverages the latent diffusion model (LDM) to enhance the efficiency and precision of MRI generation tasks. GM-LDM integrates a 3D autoencoder, pre-trained on the large-scale ABCD MRI dataset, achieving statistical consistency through KL divergence loss. We employ a Vision Transformer (ViT)-based encoder-decoder as the denoising network to optimize generation quality. The framework flexibly incorporates conditional data, such as functional network connectivity (FNC) data, enabling personalized brain imaging, biomarker identification, and functional-to-structural information translation for brain diseases like schizophrenia.