Sparsity-based acoustic inversion in cross-sectional multiscale optoacoustic imaging.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Imaging

Background:

• Advancements in optoacoustic imaging hardware enable high-resolution, single-shot image acquisition, reducing motion artifacts.
• Image reconstruction algorithms are crucial for accurate imaging from noisy, distorted projection data, addressing remaining artifacts like signal distortion and out-of-plane signals.

Purpose of the Study:

• To evaluate the effectiveness of a model-based approach combined with sparsity-based inversion (TV-L1) for optoacoustic image reconstruction.
• To compare the performance of TV-L1 inversion against conventional methods like Tikhonov regularization, TV inversion, and L1 inversion.

Main Methods:

• Utilized a model-based acoustic inversion approach incorporating L1 norm and total variation (TV) regularization.
• Employed a numerically efficient nonlinear gradient descent algorithm to solve the optimization problem.
• Tested the TV-L1 model-based inversion technique on both numerically generated and in vivo experimental data from adult mice.

Main Results:

• TV-L1 inversion demonstrated superior performance compared to Tikhonov, TV, and L1 inversion methods.
• Reconstructed images using TV-L1 inversion showed higher quantitative similarity to original images in numerical tests.
• Experimental data showed that TV-L1 inversion produced sharper images with reduced streak artifacts.

Conclusions:

• TV-L1 inversion effectively enhances the quality of multiscale optoacoustic images acquired in vivo.
• The high fidelity of model-based TV-L1 inversion positions it as a potential new standard for cross-sectional optoacoustic image reconstruction.