A photoacoustic imaging reconstruction method based on directional total variation with adaptive directivity

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Classical total variation (TV) algorithms in photoacoustic tomography (PAT) struggle with preserving image edges and textures due to their lack of directional sensitivity.
• This limitation hinders the quality of reconstructed PAT images, necessitating the development of improved algorithms.

Purpose of the Study:

• To introduce a novel directional total variation with adaptive directivity (DDTV) model for PAT image reconstruction.
• To address the shortcomings of traditional TV algorithms in edge and texture preservation.

Main Methods:

• Developed a DDTV model that weights image gradients based on adaptive, spatially varying directivity patterns.
• Employed a gradient-based approach to estimate the image's orientation field.
• Incorporated a reliability parameter for orientation field estimation to weight image gradients.
• Derived an efficient iterative algorithm with adaptive directivity updates for solving the DDTV reconstruction problem.

Main Results:

• DDTV-based PAT reconstruction significantly outperformed filtered back-projection (FBP) and TV algorithms in numerical simulations.
• Reconstructed images using DDTV showed peak signal-to-noise ratios (PSNR) approximately 10 dB higher than TV and 18 dB higher than FBP.
• In vitro experiments demonstrated DDTV's effectiveness, yielding sharper edges and clearer textures compared to TV.

Conclusions:

• The DDTV algorithm offers a substantial improvement in the quality of reconstructed PAT images, particularly for phantoms with diverse directivity patterns.
• DDTV effectively preserves image edges and texture details, overcoming limitations of classical TV methods.
• Both numerical simulations and in vitro experiments validate the robustness and effectiveness of the DDTV algorithm for PAT image reconstruction.