Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

McCaD: Multi-Contrast MRI Conditioned, Adaptive Adversarial Diffusion Model for High-Fidelity MRI Synthesis

Sanuwani Dayarathna, Kh Tohidul Islam, Bohan Zhuang, Guang Yang, Jianfei Cai, Meng Law, Zhaolin Chen

TL;DR

McCaD (Multi-Contrast MRI Conditioned Adaptive Adversarial Diffusion), a novel framework leveraging an adversarial diffusion model conditioned on multiple contrasts for high-fidelity MRI synthesis significantly enhances synthesis accuracy by employing a multi-scale, feature-guided mechanism, incorporating denoising and semantic encoders.

Abstract

Magnetic Resonance Imaging (MRI) is instrumental in clinical diagnosis, offering diverse contrasts that provide comprehensive diagnostic information. However, acquiring multiple MRI contrasts is often constrained by high costs, long scanning durations, and patient discomfort. Current synthesis methods, typically focused on single-image contrasts, fall short in capturing the collective nuances across various contrasts. Moreover, existing methods for multi-contrast MRI synthesis often fail to accurately map feature-level information across multiple imaging contrasts. We introduce McCaD (Multi-Contrast MRI Conditioned Adaptive Adversarial Diffusion), a novel framework leveraging an adversarial diffusion model conditioned on multiple contrasts for high-fidelity MRI synthesis. McCaD significantly enhances synthesis accuracy by employing a multi-scale, feature-guided mechanism, incorporating denoising and semantic encoders. An adaptive feature maximization strategy and a spatial feature-attentive loss have been introduced to capture more intrinsic features across multiple contrasts. This facilitates a precise and comprehensive feature-guided denoising process. Extensive experiments on tumor and healthy multi-contrast MRI datasets demonstrated that the McCaD outperforms state-of-the-art baselines quantitively and qualitatively. The code is provided with supplementary materials.

McCaD: Multi-Contrast MRI Conditioned, Adaptive Adversarial Diffusion Model for High-Fidelity MRI Synthesis

TL;DR

McCaD (Multi-Contrast MRI Conditioned Adaptive Adversarial Diffusion), a novel framework leveraging an adversarial diffusion model conditioned on multiple contrasts for high-fidelity MRI synthesis significantly enhances synthesis accuracy by employing a multi-scale, feature-guided mechanism, incorporating denoising and semantic encoders.

Abstract

Magnetic Resonance Imaging (MRI) is instrumental in clinical diagnosis, offering diverse contrasts that provide comprehensive diagnostic information. However, acquiring multiple MRI contrasts is often constrained by high costs, long scanning durations, and patient discomfort. Current synthesis methods, typically focused on single-image contrasts, fall short in capturing the collective nuances across various contrasts. Moreover, existing methods for multi-contrast MRI synthesis often fail to accurately map feature-level information across multiple imaging contrasts. We introduce McCaD (Multi-Contrast MRI Conditioned Adaptive Adversarial Diffusion), a novel framework leveraging an adversarial diffusion model conditioned on multiple contrasts for high-fidelity MRI synthesis. McCaD significantly enhances synthesis accuracy by employing a multi-scale, feature-guided mechanism, incorporating denoising and semantic encoders. An adaptive feature maximization strategy and a spatial feature-attentive loss have been introduced to capture more intrinsic features across multiple contrasts. This facilitates a precise and comprehensive feature-guided denoising process. Extensive experiments on tumor and healthy multi-contrast MRI datasets demonstrated that the McCaD outperforms state-of-the-art baselines quantitively and qualitatively. The code is provided with supplementary materials.
Paper Structure (15 sections, 12 equations, 4 figures, 8 tables)

This paper contains 15 sections, 12 equations, 4 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Semantic Encoder (SE)
  5. Diffusive Encoder (DE)
  6. Adaptive Feature Maximizer (FM)
  7. Feature Attentive (FA) Loss Function
  8. Model Architecture
  9. Experiments and Results
  10. Datasets and Implementation Details
  11. Competing Methods and Evaluation Metrics
  12. Performance Comparison
  13. Tumor Segmentation
  14. Ablation Studies
  15. Conclusion

Figures (4)

  • Figure 1: Network architecture of McCaD. A: Overall Architecture, B: Multi-scale Feature Guided Denoising Network to incorporate feature characteristics from conditional MRI contrasts at various stages to guide the reverse diffusion process, C: Adaptive Feature Maximizer, to weights more pertinent features within the latent space D: Feature Attentive Loss to improve the perceptual quality of the synthetic results.
  • Figure 2: Visualization of results for T2w and T1w Healthy brain MRI synthesis with corresponding error maps. McCaD yields lower artifacts with higher anatomical fidelity compared to baselines.
  • Figure 3: Visualization of results for T2w and FLAIR tumor brain MRI synthesis on BraTS dataset with corresponding error maps. McCaD shows improved synthetic results compared to baselines with fewer errors and more accurate depictions of pathological regions.
  • Figure 4: Visualization of tumor segmentation results. Column (A) shows the tumor mask segmented using ground truths. Using different methods, the first row (B-F) shows the tumor segmentation masks with T1w, synthetic FLAIR, and synthetic T2w. The second row (B-F) shows tumor segmentation masks from T1w, FLAIR, and synthetic T2w.