Compressed system models in multispectral optoacoustic tomography

One of the challenges of multispectral optoacoustic tomography (MSOT) is the reconstruction of the images from the projection data. Conventionally, analytical inversion formulae are used owing to their simplicity and numerical efficiency. However, such solutions are often limited to ideal detection scenarios and lead to image artifacts when the system characteristics deviate from the assumed ones. In such cases, image quality may be improved by adopting a model-based approach in which the MSOT system is modeled via a matrix relation, which is subsequently inverted using established algebraic techniques to reconstruct the image. Nonetheless, model-based inversion is usually more computationally demanding than its analytical counterparts owing to the large size of the model matrix. In this paper, we analyze the sparsity that exists in the model matrix and show how it may be exploited for accelerating image reconstruction. In particular, a wavelet-packet framework is presented under which the size of the model matrix may be reduced.