Sparse Mixture-of-Experts for Non-Uniform Noise Reduction in MRI Images

2025-01-24

AI Summary

MRI images commonly suffer from coupled intensity inhomogeneity (bias field) and noise, posing significant challenges for denoising. To address this, we propose the first sparse-gated Mixture-of-Experts (MoE) model specifically designed for MRI denoising. Our method employs a region-aware gating mechanism that dynamically activates dedicated CNN experts tailored to local texture and noise characteristics, enabling pixel-wise adaptive modeling. By integrating sparse expert activation, multi-expert collaboration, and end-to-end trainability, the model accurately captures spatially non-stationary noise distributions while maintaining computational efficiency. Extensive experiments on both synthetic and real clinical brain MRI datasets demonstrate substantial improvements over current state-of-the-art methods. Moreover, our approach exhibits strong generalization and robustness on unseen data across diverse scanning devices and acquisition protocols—without requiring protocol-specific retraining or fine-tuning.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Multi-modal Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Magnetic Resonance Imaging (MRI) is an essential diagnostic tool in clinical settings, but its utility is often hindered by noise artifacts introduced during the imaging process.Effective denoising is critical for enhancing image quality while preserving anatomical structures. However, traditional denoising methods, which often assume uniform noise distributions, struggle to handle the non-uniform noise commonly present in MRI images. In this paper, we introduce a novel approach leveraging a sparse mixture-of-experts framework for MRI image denoising. Each expert is a specialized denoising convolutional neural network fine-tuned to target specific noise characteristics associated with different image regions. Our method demonstrates superior performance over state-of-the-art denoising techniques on both synthetic and real-world brain MRI datasets. Furthermore, we show that it generalizes effectively to unseen datasets, highlighting its robustness and adaptability.

Problem

• MRI image enhancement
• noise reduction
• structural visibility improvement

Innovation

• Sparse Mixture of Experts
• MRI Denoising
• Robust Generalization