Pelvic Registration Using Iterative Error Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pelvic registration methods in medical surgical robot vision navigation face challenges due to the complex geometry of the pelvis, particularly the spherical concave structure of the acetabulum, leading to registration errors and inaccuracies, which can result in surgical failures and accidents.
Innovation Solution
A method and device for pelvic registration that involves a multi-step process including primary, secondary, and tertiary registrations, with point deletion and error threshold-based optimizations using iterative closest point algorithms to improve registration accuracy by eliminating noise and bad point pairs, and rotation model adjustments to avoid local optima.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional ICP algorithm is used for pelvic registration, then the registration process is simple, but the algorithm gets trapped in local optimal solutions and loses optimal solution due to noise data and bad point pairs
Solution Approach 1:
The patent divides the registration process into multiple stages: primary registration using ICP algorithm, secondary registration using RANSAC algorithm for robust outlier rejection, and tertiary registration for final refinement. This segmentation allows each algorithm to perform its strengths while mitigating weaknesses, resolving the contradiction between simplicity and accuracy.
Solution Approach 2:
The patent performs preliminary actions by conducting primary registration first to establish an initial alignment, then using this as basis for secondary registration with RANSAC. This preliminary alignment reduces the search space and improves the effectiveness of subsequent algorithms, enabling accurate registration while maintaining process efficiency.
2Measurement precision
If more marker points are used for registration, then the registration accuracy may improve, but the influence of noise and bad point pairs increases
Solution Approach 1:
The patent extracts and removes bad point pairs and noise from the marker points through RANSAC algorithm in secondary registration. By identifying and eliminating outliers, the system retains only high-quality correspondence points for final registration, thus improving accuracy while reducing noise influence.
Solution Approach 2:
The patent implements feedback mechanisms by evaluating registration errors and using this information to iteratively refine the registration process. The error metrics from primary registration guide the RANSAC algorithm in secondary registration, creating a feedback loop that continuously improves accuracy while filtering out noise.
3Measurement precision
If multiple registration steps are performed, then the registration accuracy improves, but the registration time increases
Solution Approach 1:
The patent employs dynamic algorithm selection where the complexity and type of registration steps are adapted based on the quality of point clouds and registration requirements. The system dynamically switches between ICP and RANSAC algorithms depending on the stage and data quality, optimizing the balance between accuracy and time consumption.
Solution Approach 2:
The patent performs preliminary registration using computationally efficient ICP algorithm to establish initial alignment quickly. This preliminary action reduces the burden on subsequent more time-consuming algorithms, allowing multiple registration steps to be performed while minimizing total registration time.
Data Source
AI summary
Provided are a method and device for pelvic registration, a computer-readable storage medium and a processor. The method includes: registering a three-dimensional model of a pelvis with an actual pelvis according to M first marker points to obtain primary registration model and primary registration matrix; registering the primary registration model with the actual pelvis according to N second marker points to obtain secondary registration model and secondary registration matrix; deleting second marker points having first registration errors greater than an error threshold and determining remaining second marker points as third marker points; registering the primary registration model with the actual pelvis according to the plurality of third marker points to obtain tertiary registration model and tertiary registration matrix; determining a registration matrix corresponding to a less one of a mean of first registration errors and a mean of second registration errors as an optimized registration matrix.