3D MRI-US Deformable Registration for Real-Time Intervention Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image-guided interventions face challenges due to the dynamic nature of internal anatomy, which undergoes motion and deformation during procedures, limiting the integration of imaging modalities like MRI and ultrasound for real-time guidance, and current registration methods are computationally inefficient or dependent on operator skill.
Innovation Solution
A method involving pre-interventional imaging with MRI, CT, or PET, followed by real-time ultrasound, using trained deformation models to register 3D images, allowing for real-time visualization of tissue features by leveraging the strengths of different imaging modalities, such as MRI's soft tissue contrast and ultrasound's flexibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MRI is used for pre-interventional imaging to provide superior soft tissue contrast and lesion detection, then tissue contrast and lesion conspicuity are improved, but the computational burden of fusing MRI with real-time US images increases significantly
Solution Approach 1:
The patent performs MRI-US deformation model training in advance before the actual intervention procedure. The deformation model learns the complex non-linear transformations between MRI and ultrasound modalities during a preliminary training phase using paired images, so that during real-time intervention, the pre-trained model can rapidly apply deformations without requiring heavy computation at that moment.
Solution Approach 2:
The patent creates a deformed MRI image that is transformed to match the ultrasound modality's appearance characteristics. By generating this copied representation of MRI data in the ultrasound domain, the system can overlay anatomical features from MRI onto real-time US images, providing enhanced tissue contrast while maintaining real-time performance.
2Measurement precision
If analytical deformable registration is used to register pre-interventional images with real-time US images, then alignment accuracy is improved, but computation time increases significantly making it incompatible with real-time requirements
Solution Approach 1:
The deformation model is trained in advance using paired MRI-US images to learn the complex non-linear transformations. This preliminary training phase captures the deformation patterns, so that during real-time intervention, the system can rapidly apply these pre-learned transformations without performing heavy iterative optimization computations.
Solution Approach 2:
The patent replaces traditional mechanical iterative optimization algorithms with a data-driven deep learning deformation model. Instead of performing computationally intensive iterative registration during real-time intervention, the system uses a neural network that has been trained to directly predict deformation fields, substituting the mechanical optimization process with a faster inference-based approach.
3Ease of operation
If manual registration processes are used in CT-US fusion-guided procedures, then operator flexibility is maintained, but alignment accuracy decreases due to dependence on operator skill and rigid registration methods
Solution Approach 1:
The deformation model performs automatic self-alignment between MRI and ultrasound images by learning the transformation patterns from training data. The system independently computes the optimal deformation field without requiring manual intervention or operator skill, thereby achieving consistent high accuracy while maintaining ease of operation.
Solution Approach 2:
The patent transforms the rigid registration parameters used in traditional manual methods into flexible non-linear deformation fields learned by the deep learning model. This allows the system to adapt to complex anatomical variations and non-rigid transformations that manual rigid registration cannot handle, significantly improving alignment accuracy while remaining automated and easy to operate.
4Productivity
If real-time ultrasound imaging is used for its real-time capabilities and flexibility, then imaging speed and accessibility are improved, but soft tissue visualization and lesion conspicuity deteriorate
Solution Approach 1:
The patent merges the real-time imaging capability of ultrasound with the superior soft tissue contrast of MRI by deforming the MRI image to match the ultrasound modality and overlaying them. The system combines the temporal advantages of US with the anatomical detail of MRI, achieving both real-time performance and enhanced tissue visualization simultaneously.
Solution Approach 2:
The system creates a copied representation of MRI data that is transformed into the ultrasound domain. By generating this deformed MRI copy in the US modality's appearance space, the system can overlay anatomical features from MRI onto real-time US images, providing enhanced tissue contrast and lesion conspicuity while maintaining the real-time imaging speed of ultrasound.
Data Source
AI summary
Systems and methods are provided for real-time multimodal deformable image registration for image-guided interventions. A pre-interventional three-dimensional (3D) magnetic resonance imaging (MRI) image and multiple 3D ultrasound (US) images capturing various respiratory states and poses are acquired for a patient. The MRI image is registered to each US image using a trained MR-US deformation model, producing deformed MRI images. During intervention, an interventional 3D US image is acquired and registered to a pre-interventional US image using a trained US-US deformation model, determining a warp field. This warp field is applied to the corresponding deformed MRI image, producing a registered 3D MRI image for visualizing annotated tissue features from the pre-interventional MRI on the live interventional US image. The disclosed approach leverages multimodal imaging and deep learning models to enhance visualization during interventions by combining superior soft tissue contrast of MRI with real-time US imaging capabilities.