Dental DVT Geometric Calibration with ML-Guided Parameter Estimation
Solution Overview
Problem
Conventional geometric calibration methods for dental volume tomography (DVT) images suffer from complex optimization problems, long computation times, and suboptimal results due to numerous parameters and local optima, while machine learning (ML) methods offer faster corrections but lack control over image data changes.
Innovation Solution
A combined approach that integrates conventional parameter estimation methods with ML correction methods, utilizing ML to generate data-based prior knowledge that stabilizes and accelerates conventional parameter estimation, enabling efficient compensation of motion artifacts in DVT images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional parameter estimation methods are used for geometric calibration, then raw data fidelity can be maintained and controlled, but computation time increases and results become suboptimal due to complex optimization problems with many local optima
Solution Approach 1:
The patent applies preliminary action by using a machine learning model to generate an initial estimate of geometric parameters before the conventional optimization process. This preliminary estimation provides a starting point that is closer to the global optimum, reducing the iteration time and computational burden of subsequent optimization steps while maintaining measurement precision.
2Productivity
If machine learning correction methods are used, then computation time is significantly reduced and global image relationships are considered, but control over image data changes becomes difficult and boundary conditions are hard to integrate
Solution Approach 1:
The patent uses an intermediary approach where the machine learning model serves as a preliminary estimator rather than a direct corrector. The ML model generates initial parameter estimates that are then refined by conventional optimization methods, allowing boundary conditions and control constraints to be applied in the refinement stage while benefiting from the speed of ML-based initial estimation.
3Manufacturing precision
If conventional parameter estimation methods are used, then optimization can be performed with multiple parameters and dependencies, but the complexity of the optimization problem increases with numerous local optima
Solution Approach 1:
By performing preliminary action with the machine learning model to generate initial parameter estimates, the patent simplifies the subsequent optimization problem. The complex multi-parameter optimization is reduced to a refinement process starting from a pre-estimated solution, reducing the search space and number of local optima that need to be navigated while maintaining calibration accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for the geometric calibration of a CBCT scan by updating the geometric parameters used in the reconstruction method, wherein the update of the geometric parameters is supported by a first correction method based on machine learning (ML) by using the result of the first correction method as a reference for a second correction method for parameter estimation, and wherein the second correction method for parameter estimation incorporates the measurement data of the CBCT scan, comprising the following steps: (S1) providing the measurement data of the CBCT scan and the geometric parameters; (S2) providing a first volume by applying a reconstruction method to the provided measurement data and the geometric parameters; (S3) providing a corrected volume by applying the first correction method to the first volume;(S4) Provision of updated geometric parameters by applying the second correction method to the measurement data and the corrected volume.;