Performance comparison of commonly used photoacoustic tomography reconstruction algorithms under various blurring conditions

Abstract

A reconstructed image in photoacoustic tomography (PAT) imaging modality can be blurred because of some system-dependent effects – broad laser pulse, the limited bandwidth of the sensors, acoustically attenuating medium and transducers with finite-size apertures having Gaussian sensitivity profile. A robust deblurring method involving a system matrix is presented in this work. Image reconstruction was accomplished by using the backprojection (BP), interpolated time reversal (ITR), Tikhonov regularization (TH) and total variation minimization (TV) schemes. A modified version of the BP algorithm was also explored for image reconstruction for finite transducers. Each factor of blurring was considered at a time. For the model-based methods, the system matrix was constructed in presence of such a factor. Subsequently, the model matrix inversion method was deployed for image formation. Quantitative values of some standard metrics were computed to assess the performance of the reconstruction protocols. The visual inspection of the reconstructed images as well as numerical values of the figures of merit revealed that the BP and ITR techniques generally fail to remove blurring; however, the model-based methods can provide distortion-free images under various conditions. For example, the numerical values of the Pearson correlation coefficient (PCC) for all the cases showed a strong correlation between the nominal and reconstructed images for the TH and TV techniques compared to the BP and ITR methods. The TV method can be utilized in practice to facilitate the PAT image reconstruction free from system-dependent blurring.