Path and Bone-Contour Regularized Unpaired MRI-to-CT Translation

2025-05-06

AI Summary

Existing unpaired MRI-to-CT translation methods struggle to accurately reconstruct anatomical structures—such as bone—that exhibit low contrast in MRI but high conspicuity in CT, limiting their clinical utility in radiotherapy planning. To address this, we propose a Neural Ordinary Differential Equation (Neural ODE)-based cross-modal mapping framework. Our method introduces a novel co-optimization mechanism that jointly minimizes path length and enforces learnable bone contour regularization, thereby constraining both trajectory smoothness and geometric consistency of bone boundaries under unpaired supervision. The architecture integrates shared latent-space projection, bone-region attention guidance, and a trainable bone-contour generation network. Evaluated on three public datasets, our approach significantly reduces CT reconstruction error; downstream bone segmentation achieves Dice improvements of 3.2–5.8%, and bone boundary localization accuracy reaches sub-pixel precision. This work establishes a new high-fidelity paradigm for unsupervised bone modeling in medical image translation.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Accurate MRI-to-CT translation promises the integration of complementary imaging information without the need for additional imaging sessions. Given the practical challenges associated with acquiring paired MRI and CT scans, the development of robust methods capable of leveraging unpaired datasets is essential for advancing the MRI-to-CT translation. Current unpaired MRI-to-CT translation methods, which predominantly rely on cycle consistency and contrastive learning frameworks, frequently encounter challenges in accurately translating anatomical features that are highly discernible on CT but less distinguishable on MRI, such as bone structures. This limitation renders these approaches less suitable for applications in radiation therapy, where precise bone representation is essential for accurate treatment planning. To address this challenge, we propose a path- and bone-contour regularized approach for unpaired MRI-to-CT translation. In our method, MRI and CT images are projected to a shared latent space, where the MRI-to-CT mapping is modeled as a continuous flow governed by neural ordinary differential equations. The optimal mapping is obtained by minimizing the transition path length of the flow. To enhance the accuracy of translated bone structures, we introduce a trainable neural network to generate bone contours from MRI and implement mechanisms to directly and indirectly encourage the model to focus on bone contours and their adjacent regions. Evaluations conducted on three datasets demonstrate that our method outperforms existing unpaired MRI-to-CT translation approaches, achieving lower overall error rates. Moreover, in a downstream bone segmentation task, our approach exhibits superior performance in preserving the fidelity of bone structures. Our code is available at: https://github.com/kennysyp/PaBoT.

Problem

• Improving MRI-to-CT translation for unpaired datasets
• Enhancing bone structure accuracy in translated images
• Optimizing treatment planning in radiation therapy

Innovation

• Path-regularized MRI-to-CT translation using neural ODEs
• Bone-contour guided mapping for anatomical accuracy
• Shared latent space projection for unpaired datasets