Spatially variant regularization based on model resolution and fidelity embedding characteristics improves...

Area of Science:

• Medical Imaging
• Biomedical Optics
• Computational Imaging

Background:

• Photoacoustic tomography (PAT) reconstruction is an ill-posed inverse problem, often requiring regularization like Tikhonov (ST) or total variation (TV) for meaningful results.
• Existing regularization methods do not adequately address non-uniform detector sensitivity and system parameter variations inherent in PAT.
• Limited boundary data in PAT exacerbates reconstruction challenges, necessitating improved regularization strategies.

Discussion:

• Two novel regularization schemes, model resolution (spatially varying) and fidelity-embedded (orthogonality-based), were developed within the Tikhonov framework.
• These methods explicitly account for spatially varying sensitivity and system matrix properties in photoacoustic imaging.
• Systematic evaluation using numerical simulations and in-vivo mouse data demonstrated the efficacy of the proposed schemes.

Key Insights:

• The proposed spatially varying regularization schemes significantly outperform standard Tikhonov and total variation methods in PAT.
• Improvements include at least a 2 dB increase in signal-to-noise ratio (SNR), representing a 1.58-fold enhancement.
• These advancements enable more accurate reconstruction of initial pressure rise distributions from limited, noisy photoacoustic data.

Outlook:

• Future research could explore further optimization of spatially varying regularization parameters for diverse PAT applications.
• Integration of these advanced regularization techniques may enhance diagnostic capabilities in preclinical and clinical settings.
• These methods hold promise for improving the resolution and quantitative accuracy of photoacoustic tomography.