Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Konstantinos Barmpounakis, Theodoros P. Vagenas, Maria Vakalopoulou, George K. Matsopoulos

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.

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

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.
Paper Structure (12 sections, 4 equations, 2 figures, 1 table)

This paper contains 12 sections, 4 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methodology
  3. Architectures
  4. Loss function
  5. Quantitative Evaluation
  6. Experiments
  7. Dataset & Preprocessing
  8. Implementation Details
  9. Results
  10. Conclusion

Figures (2)

  • Figure 1: Illustration of the overall proposed framework. Input MRI volumes undergo Z-score normalization and feed a Mamba-based encoder–decoder architecture. The predicted output is transformed back to HU via reverse normalization on dataset-level and appropriate clipping. During training, the network is supervised using paired 3D CT volumes with a staged compound loss function
  • Figure 2: Visual comparison of U-Net, SwinUNETR, nnU-Net, SegMamba, and U-Mamba on representative cases for each anatomical site. The target CT images and corresponding MRI images are shown in the last columns