Mamba-driven MRI-to-CT Synthesis for MRI-only Radiotherapy Planning

2026-03-24

AI Summary

This work proposes the first method to apply the Mamba architecture to MRI-to-CT synthesis, aiming to enable radiation therapy planning using MRI alone and thereby eliminate ionizing radiation exposure and multimodal registration errors. By integrating 3D state space models from U-Mamba and SegMamba, the approach effectively captures long-range dependencies to generate high-quality volumetric CT images. Evaluated on the SynthRAD2025 dataset, the method achieves excellent image similarity and anatomical consistency, while demonstrating high accuracy in Hounsfield unit quantification and geometric fidelity as assessed by TotalSegmentator. Furthermore, it exhibits fast inference speed, highlighting the promising potential of Mamba-based models in clinical radiotherapy workflows.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

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

Abstract

Radiotherapy workflows for oncological patients increasingly rely on multi-modal medical imaging, commonly involving both Magnetic Resonance Imaging (MRI) and Computed Tomography (CT). MRI-only treatment planning has emerged as an attractive alternative, as it reduces patient exposure to ionizing radiation and avoids errors introduced by inter-modality registration. While nnU-Net-based frameworks are predominantly used for MRI-to-CT synthesis, we explore Mamba-based architectures for this task, aiming to showcase the advantages of state-space modeling for cross-modality translation compared to standard convolutional neural networks. Specifically, we adapt both the U-Mamba and the SegMamba architecture, originally proposed for segmentation, to perform cross-modality image generation. Our 3D Mamba architecture effectively captures complex volumetric features and long-range dependencies, thus allowing accurate CT synthesis while maintaining fast inference times. Experiments were conducted on a subset of SynthRAD2025 dataset, comprising registered single-channel MRI-CT volume pairs across three anatomical regions. Quantitative evaluation is performed via a combination of image similarity metrics computed in Hounsefield Units (HU) and segmentation-based metrics obtained from TotalSegmentator to ensure geometric consistency is preserved. The findings pave the way for the integration of state-space models into radiotherapy workflows.

Problem

• MRI-to-CT synthesis
• MRI-only radiotherapy planning
• cross-modality translation
• medical image synthesis

Innovation

• Mamba
• MRI-to-CT synthesis
• state-space model
• radiotherapy planning
• cross-modality translation