CT Geometric Calibration via Learning Model Probability Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current geometric calibration methods for cone-beam CT devices fail to accurately obtain imaging direction information due to reliance on parameterized representative values and complex phantom manufacturing and verification processes.
Innovation Solution
A geometric calibration method and apparatus that uses a learning model to detect points from projection regions, calculate probability distributions, extract samples, and determine candidate projection matrices by transforming correspondences between markers and points, designating the matrix as valid if the difference between transformed and detected points is within a threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If parameterized representative values are used for geometric calibration, then the calibration process is simplified, but the imaging direction information cannot be accurately obtained
Solution Approach 1:
The patent introduces a learning model as an intermediary between the projection regions and the marker correspondences. The learning model processes the complex relationship between detected points and markers, enabling accurate imaging direction calculation without requiring complex manual parameterization. This intermediary computational layer resolves the contradiction by automating the precise matching process that would otherwise require simplified but inaccurate representative values.
2Measurement precision
If geometric calibration phantom with markers of different sizes is used, then imaging direction information can be obtained, but manufacturing and verification become complicated
Solution Approach 1:
The patent changes the approach from using physical phantom markers of different sizes to using a learning model that processes numerical data from projection regions. Instead of varying physical parameters (marker sizes), the system varies computational parameters (probability distributions, sample selections) to achieve the same calibration objective. This eliminates the manufacturing complexity while maintaining the ability to obtain accurate imaging direction information.
3Measurement precision
If learning model with probability distribution and sample extraction is used, then accurate imaging direction information is obtained, but computational complexity increases
Solution Approach 1:
The patent applies partial action by extracting a predetermined number of samples from the probability distribution rather than processing all possible marker-point correspondences. This sampling approach achieves sufficient calibration accuracy without the excessive computational burden of exhaustive processing. The selective sampling provides a practical balance between precision and computational complexity.
Data Source
AI summary
A geometric calibration apparatus detects points from projection regions onto which markers disposed on a phantom are projected, and calculates an output vector representing a probability distribution that gives a probability with which each point is a projection of each marker, by inputting data corresponding to each point to a learning model. The geometric calibration apparatus extracts a predetermined number of samples based on the probability distribution, obtains a candidate projection matrix by transforming correspondences between markers determined based on the samples among the markers and points determined based on the samples among the points, calculates points into which the markers are transformed by the candidate projection matrix, calculates a difference between a set of the transformed points and a set of the detected points, and designates the candidate projection matrix as a projection matrix when the difference is less than or equal to a threshold.