Atlas Registration with Simulated Tumor Growth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Elastic registration of patient images containing pathologies, such as tumors, to an atlas is challenging due to the absence of the pathology in the atlas, leading to unsatisfactory results from standard algorithms, especially for large tumor masses.
Innovation Solution
A data processing method that simulates the growth of pathological structures within an atlas by using meta-information and elastic image fusion, allowing for the registration of patient data containing specific structures to atlas data without such structures, through iterative steps of segmentation, fusion, and simulation, utilizing meta-data for accurate tumor modeling and tissue interaction rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard elastic registration algorithms are used to register patient images containing tumors to an atlas, then the registration process is simple and fast, but the registration accuracy is unsatisfactory due to the mass effect of tumors not being accounted for
Solution Approach 1:
The patent applies preliminary action by simulating the tumor mass effect on the atlas image before performing the actual registration. A seed point is placed at the tumor center and a radial growth model is used to simulate the tumor's mass effect, creating a pre-modified atlas that accounts for the expected deformation. This preliminary simulation prepares the atlas in advance to match the patient's actual anatomy, thereby improving registration accuracy without requiring complex iterative adjustments during the registration process itself.
2Measurement precision
If biomechanical models of soft tissue deformation are used to simulate tumor mass effect, then the simulation accuracy is high, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent employs a simplified radial growth model instead of complex biomechanical models. The simulation uses a straightforward mathematical approach where the tumor is modeled as a radial expansion from a seed point, with deformation magnitude proportional to the distance from the tumor center. This simplified model achieves sufficient accuracy for clinical purposes while being computationally efficient and fast to execute, effectively replacing expensive complex models with a cheaper, faster alternative that meets the required performance level.
3Measurement precision
If the atlas is modified to include simulated tumor structures, then the registration of pathological patient images improves, but the atlas becomes more complex and requires additional simulation parameters
Solution Approach 1:
The patent applies local quality by modifying only the local region of the atlas where the tumor is located, rather than transforming the entire atlas. The radial growth model applies deformation locally around the tumor seed point, affecting only the surrounding tissue in proportion to its distance from the tumor center. This localized approach improves registration accuracy for pathological regions while maintaining the original atlas structure elsewhere, thereby minimizing the increase in overall atlas complexity.
Data Source
Figure 1
Figure 2
AI summary
A data processing method for registering patient data containing normal patient data and specific object data to atlas data not containing this specific object data, the method comprising the following steps performed by a computer: a) acquiring the patient data which comprises anatomical information of the patient including normal patient data and the specific object data of the patient; b) segmenting the specific object data in the patient data; c) acquiring atlas data; d) acquiring meta-information about components of the atlas data; e) performing a fusion of the acquired patient data excluding the segmented specific object data and the atlas data to obtain registered atlas data; and f) simulating the growth of a simulated specific object within the registered atlas data using the registered atlas data and the meta- information about components of the atlas data being adjacent to the area of the specific object to obtain registered atlas data containing a simulated grown specific object.