Coded Aperture Optimization in X-Ray Computed Tomography via Sparse Covariance Matrix Estimation.

Area of Science:

• Medical Imaging
• Computational Imaging
• Applied Mathematics

Background:

• Coded aperture X-ray computed tomography (CAXCT) reconstructs object interiors from coded projections.
• Current CAXCT methods are computationally intensive, limiting image size and reconstruction quality.
• Optimizing coded apertures is crucial for improving CAXCT performance.

Purpose of the Study:

• To develop a computationally efficient method for minimizing information loss in CAXCT projections.
• To enable the reconstruction of large tomographic images using optimized coded apertures.
• To enhance the quality of reconstructed images in CAXCT.

Main Methods:

• Introduced a sparse covariance matrix estimation approach to quantify projection information overlap.
• Utilized block matrix multiplication and sparse estimation to derive the covariance matrix.
• Developed a heuristic algorithm with a noise factor to identify optimal combinations of coded aperture elements (D) for maximum non-overlapping information.

Main Results:

• The proposed sparse covariance matrix method demonstrated optimization performance comparable to state-of-the-art techniques on simulated datasets.
• Successfully performed coded aperture optimization for large 512 × 512 images.
• Analysis of covariance matrix coefficients revealed sparsity, with significant information captured by analyzing <0.02% of coefficients.

Conclusions:

• The sparse covariance matrix estimation approach effectively minimizes information loss in CAXCT, enabling large image reconstruction.
• The method offers a computationally efficient alternative for optimizing coded apertures in X-ray computed tomography.
• This work advances CAXCT by providing a scalable solution for high-resolution image reconstruction.