MoEDiff-SR: Mixture of Experts-Guided Diffusion Model for Region-Adaptive MRI Super-Resolution

2025-04-09

AI Summary

To address the clinical and research bottleneck of low spatial resolution in 3T MRI—hindering visualization of fine brain structures (e.g., centrum semiovale, cortical gyri/sulci, gray–white matter boundaries)—this paper proposes a region-adaptive super-resolution method. We pioneer the integration of a Mixture-of-Experts (MoE) mechanism into a diffusion model, synergizing Transformer-based multi-scale feature extraction with token-level gated routing to enable brain-region-specific dynamic denoising and interpretable reconstruction. A probabilistic weighted fusion strategy is further introduced to enhance reconstruction consistency. Quantitatively, perceptually, and computationally, our method outperforms state-of-the-art approaches across all metrics. Clinical evaluation demonstrates significantly improved detection of subtle lesions, while difference-map analysis validates both the anatomical specificity and rational expert specialization inherent in our MoE design.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Magnetic Resonance Imaging (MRI) at lower field strengths (e.g., 3T) suffers from limited spatial resolution, making it challenging to capture fine anatomical details essential for clinical diagnosis and neuroimaging research. To overcome this limitation, we propose MoEDiff-SR, a Mixture of Experts (MoE)-guided diffusion model for region-adaptive MRI Super-Resolution (SR). Unlike conventional diffusion-based SR models that apply a uniform denoising process across the entire image, MoEDiff-SR dynamically selects specialized denoising experts at a fine-grained token level, ensuring region-specific adaptation and enhanced SR performance. Specifically, our approach first employs a Transformer-based feature extractor to compute multi-scale patch embeddings, capturing both global structural information and local texture details. The extracted feature embeddings are then fed into an MoE gating network, which assigns adaptive weights to multiple diffusion-based denoisers, each specializing in different brain MRI characteristics, such as centrum semiovale, sulcal and gyral cortex, and grey-white matter junction. The final output is produced by aggregating the denoised results from these specialized experts according to dynamically assigned gating probabilities. Experimental results demonstrate that MoEDiff-SR outperforms existing state-of-the-art methods in terms of quantitative image quality metrics, perceptual fidelity, and computational efficiency. Difference maps from each expert further highlight their distinct specializations, confirming the effective region-specific denoising capability and the interpretability of expert contributions. Additionally, clinical evaluation validates its superior diagnostic capability in identifying subtle pathological features, emphasizing its practical relevance in clinical neuroimaging. Our code is available at https://github.com/ZWang78/MoEDiff-SR.

Problem

• Enhancing low-resolution MRI for better clinical diagnosis
• Adaptive super-resolution for region-specific MRI details
• Improving image quality and diagnostic accuracy in neuroimaging

Innovation

• Mixture of Experts-guided diffusion model
• Region-adaptive MRI super-resolution
• Transformer-based feature extractor