Brain Imaging-to-Graph Generation using Adversarial Hierarchical Diffusion Models for MCI Causality Analysis

TL;DR

This work targets the extraction of brain effective connectivity from fMRI to aid mild cognitive impairment analysis. It introduces BIGG, a unified framework that combines diffusion denoising probabilistic models with conditional GANs and hierarchical transformers to translate fMRI signals into causal brain graphs efficiently and with high fidelity. The model employs a multi-resolution discriminator and a suite of hybrid losses to ensure realistic, diverse BECs while enforcing sparsity and enabling disease classification, validated on the ADNI dataset with superior performance and clinically plausible connectivity changes. By providing end-to-end generation of subject-specific effective connectivity and identifying MCI-related causal patterns, BIGG offers a promising tool for early diagnosis and biomarker discovery in cognitive disorders.

Abstract

Effective connectivity can describe the causal patterns among brain regions. These patterns have the potential to reveal the pathological mechanism and promote early diagnosis and effective drug development for cognitive disease. However, the current methods utilize software toolkits to extract empirical features from brain imaging to estimate effective connectivity. These methods heavily rely on manual parameter settings and may result in large errors during effective connectivity estimation. In this paper, a novel brain imaging-to-graph generation (BIGG) framework is proposed to map functional magnetic resonance imaging (fMRI) into effective connectivity for mild cognitive impairment (MCI) analysis. To be specific, the proposed BIGG framework is based on the diffusion denoising probabilistic models (DDPM), where each denoising step is modeled as a generative adversarial network (GAN) to progressively translate the noise and conditional fMRI to effective connectivity. The hierarchical transformers in the generator are designed to estimate the noise at multiple scales. Each scale concentrates on both spatial and temporal information between brain regions, enabling good quality in noise removal and better inference of causal relations. Meanwhile, the transformer-based discriminator constrains the generator to further capture global and local patterns for improving high-quality and diversity generation. By introducing the diffusive factor, the denoising inference with a large sampling step size is more efficient and can maintain high-quality results for effective connectivity generation. Evaluations of the ADNI dataset demonstrate the feasibility and efficacy of the proposed model. The proposed model not only achieves superior prediction performance compared with other competing methods but also predicts MCI-related causal connections that are consistent with clinical studies.

Figures (12)

• Figure 1: The architecture of the proposed BIGG framework. In the forward process, the empirical sample $F_0$ is transformed to the normal Gaussian noisy sample $F_T$ by gradually adding noise. In the reverse process, the fMRI is considered as a condition to guide the network $p_ \theta$ to remove the noise from $F_T$.
• Figure 2: Detailed network structure of one denoising step. It includes one generator, one discriminator, and one classifier. The input is the conditional fMRI and noisy sample $F_t$, and the output is the denosed sample $F_{t-s}'$ and BEC.
• Figure 3: The structure of the multi-channel adaptor. The inputs are the noisy sample $\mathbf{F}_t$ and the rough sample $\mathbf{X}$, and the output is the fused sample. The three samples share the same dimension.
• Figure 4: The detailed structure of the SeTe block. The input and output share the same dimension. It passes successively through the spatial multi-head attention module and the temporal multi-head attention module.
• Figure 5: The structure of the BEC estimator. The outputs are a denoised sample and an asymmetric brain connectivity matrix.