Efficient framework for model-based tomographic image reconstruction using wavelet packets.

Area of Science:

• Medical Imaging
• Computational Science
• Signal Processing

Background:

• Model-based algorithms offer advantages in tomographic imaging but face computational complexity challenges.
• Large forward-model matrices in high-resolution imaging hinder efficient reconstruction.
• Existing methods struggle with the computational demands of large-scale tomographic reconstructions.

Purpose of the Study:

• To develop a novel framework for efficient model-based tomographic reconstructions.
• To address the computational complexity associated with large forward-model matrices.
• To improve the feasibility of advanced inversion algorithms in practical imaging scenarios.

Main Methods:

• Utilized a wavelet-packet representation for both the imaged object and projection data.
• Leveraged the frequency localization property of wavelet packets to create an approximately separable model matrix.
• Separated the large model matrix into smaller, independent matrices for each spatial frequency band.

Main Results:

• Demonstrated that reconstruction at each spatial frequency band is independent and requires only a fraction of the projection data.
• Successfully reduced the effective size of the model matrix, enabling nonlinear inversion schemes.
• Validated the methodology's performance using 2-D optoacoustic tomography with both simulated and experimental data.

Conclusions:

• The proposed wavelet-packet framework offers a computationally efficient solution for model-based tomographic reconstructions.
• This approach facilitates the use of complex inversion algorithms by managing large matrix sizes.
• The method shows promise for advancing various tomographic imaging modalities, including optoacoustic tomography.