A convolutional neural network approach to calibrating the rotation axis for X-ray computed tomography

Area of Science:

• Medical Imaging
• Computational Science
• Materials Science

Background:

• Accurate calibration of the center-of-rotation is crucial for high-quality X-ray tomography reconstruction.
• Conventional methods for center-of-rotation calibration can be time-consuming and sensitive to noise and artifacts.
• Machine learning offers a promising avenue for developing more robust and automated calibration techniques.

Purpose of the Study:

• To develop and validate a machine learning algorithm for precise center-of-rotation calibration in X-ray tomography.
• To assess the accuracy and robustness of the proposed algorithm using both synthetic and experimental datasets.
• To provide an open-source tool for integrating this calibration method into existing workflows.

Main Methods:

• Implementation of a Convolutional Neural Network (CNN) for center-of-rotation calibration.
• Evaluation using synthetic X-ray tomography data with varying noise levels.
• Validation with experimental data from four distinct shale samples acquired at synchrotron facilities.

Main Results:

• The CNN-based algorithm demonstrated excellent accuracy in synthetic data evaluations across different noise ratios.
• Experimental validation showed results comparable to visual inspection and superior robustness compared to conventional methods.
• The algorithm effectively calibrated the center-of-rotation for diverse geological samples.

Conclusions:

• Convolutional Neural Networks provide an accurate and robust solution for X-ray tomography center-of-rotation calibration.
• The developed method holds potential for mitigating various imaging artifacts beyond center-of-rotation errors.
• An open-source toolbox (xlearn) is available, facilitating easy integration into synchrotron imaging pipelines.