Pseudo-CT Generation via Patch-Based Atlas Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for transforming magnetic resonance (MR) images into computed tomography (CT) images, known as pseudo-CT generation, are limited by registration errors between atlas images, which affect the accuracy of derived CT images and require additional radiation exposure and imaging time.
Innovation Solution
A method that compares patches of MR images with corresponding patches in registered atlas MR and CT images to compute similarity indicators, weight factors, and generates derived CT images through weighted combinations, accommodating registration errors and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a simplified transformation technique is used to map points directly between registered atlas images, then the process is simple and fast, but the accuracy deteriorates due to registration errors
Solution Approach 1:
The patent divides the image into multiple patches instead of processing individual points. Each patch is compared with corresponding patches in the atlas images to compute similarity indicators, which are then used to determine weight factors for combining atlas data points. This segmentation approach allows the system to accommodate registration errors while maintaining processing efficiency.
Solution Approach 2:
The patent introduces similarity indicators and weight factors as intermediate parameters to bridge the simplified mapping approach and the accuracy requirement. By computing weight factors based on patch similarities and using weighted combinations of atlas data points, the system transforms the rigid point-to-point mapping into a flexible parameter-based transformation that accounts for registration errors.
2Measurement precision
If additional CT imaging is performed to obtain accurate CT characteristics, then the measurement precision improves, but the harmful factors increase due to radiation exposure and imaging time
Solution Approach 1:
The patent creates a pseudo-CT image by copying and transforming MR image data through weighted combinations with atlas CT data. Instead of acquiring actual CT radiation data, the system generates a synthetic CT image that preserves the necessary CT characteristics (Hounsfield units, tissue density information) by leveraging the registered atlas CT image and adjusting it based on the patient's specific MR image characteristics.
Solution Approach 2:
The patent uses the registered atlas CT image as an intermediary to bridge the MR image and the desired CT characteristics. The atlas CT image serves as a template that contains accurate CT data, and the system uses similarity indicators and weight factors to adapt this template to the specific patient case, thereby obtaining accurate CT characteristics without requiring additional radiation exposure.
Data Source
AI summary
Disclosed herein are techniques for translating a first image of a first type to a derived image of a second type. For example, a plurality of similarity indicators can be computed as between a plurality of patches of first image and a plurality of patches of a first atlas image, the first atlas image being of the first type. Weight factors can then be computed based on the computed similarity indicators. These weight factors can be applied to a plurality of data points of a second atlas image to compute a plurality of data points for the derived image such that each of at least a plurality of the data points for the derived image is a function of a plurality of the data points of the second atlas image. Such a technique can be used for pseudo-CT generation from an MR image.