Non-Linear Brain Deformation Registration for MR-to-CT Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to effectively register pre-operative magnetic resonance images with intra-operative CBCT/CT images due to brain-shift, which is not addressed by MRI availability in most hospitals, necessitating a method to merge information for accurate neurosurgical planning.
Innovation Solution
A computer-implemented method for predicting a registration between medical images by obtaining a model-based segmentation of anatomical structures and determining a dense deformation field to align pre-operative MR images with intra-operative CBCT/CT images, accounting for non-linear brain deformations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pre-operative MR images are registered to intra-operative CBCT/CT images using conventional methods, then the registration accuracy deteriorates due to brain-shift, but MR images provide superior tissue contrast for pathology assessment
Solution Approach 1:
The patent introduces an intermediary registration method that transforms pre-operative MR images to CT space using a deformation field derived from both MR and CT images. This intermediary transformation enables accurate registration between MR and CBCT/CT images by accounting for brain-shift through a two-step process: first registering MR to CT, then applying the deformation field to align with intra-operative CBCT/CT images.
Solution Approach 2:
The patent performs preliminary registration of pre-operative MR images to pre-operative CT images before the surgical procedure to establish a deformation field that captures brain-shift. This preliminary action creates a transformation model that can be applied during surgery to correct for brain-shift in real-time, improving registration accuracy without requiring intra-operative MR imaging.
2Measurement precision
If intra-operative MR imaging is used to maintain registration accuracy, then measurement precision improves, but the cost and availability worsen due to MRI not being available in most hospitals
Solution Approach 1:
The patent creates a virtual copy of the pre-operative MR image data in the CT/CBCT coordinate system by applying a deformation field. This virtual copy allows the benefits of high-contrast MR imaging to be utilized in standard CT/CBCT workflows without requiring actual intra-operative MR scanners, thereby improving availability while maintaining registration accuracy.
Solution Approach 2:
The patent replaces the need for physical intra-operative MR imaging hardware with a computational approach using deformation fields and image registration algorithms. This substitution eliminates the requirement for expensive and space-consuming MR scanners in the operating room, making the system widely available in hospitals without such equipment while achieving comparable registration accuracy.
3Device complexity
If anatomical segmentations are registered directly between pre-operative MR and intra-operative CBCT/CT images, then the process is simple, but the registration accuracy deteriorates due to non-linear brain deformations
Solution Approach 1:
The patent segments the brain into different anatomical structures and applies the deformation field specifically to these segmented regions. This allows the registration to account for non-linear brain deformations in a structured manner, improving accuracy by treating different anatomical regions appropriately while maintaining computational efficiency.
Data Source
Figure 1~2
Figure 3A~3B
Figure 4
AI summary
Proposed concepts thus aim to provide schemes, solutions, concept, designs, methods and systems pertaining to predicting a registration between two medical images of a subject's brain. In particular, embodiments aim to provide a method for predicting a registration between two medical images of a subject's brain. This can be achieved by obtaining a segmentation of a first medical image to identify an anatomical structure and then, based on this, determining a dense deformation field from the first medical image to the second medical image. In other words, it is proposed that by determining a dense deformation field between a first medical image of a subject's brain and a second medical image in which a subject's brain is non-linearly deformed compared to the brain in the first medical image (i.e., has undergone brain-shift), a registration between the two images can essentially be predicted.