Model-based image reconstruction in optoacoustic tomography

The effect of limited tomographic coverage on the quality of the reconstructed image [13]. (a) A reconstruction of a cylindrical phantom with a square inclusion from full-view experimental data. (b) Model-based reconstruction with no regularizarion when only 130º of the projection data is used. (c) Back-projection reconstruction of the phantom in the limited-view scenario. (d) Regularized model-based reconstruction of the phantom in the limited-view scenario.

source:© 2014 Computer Vision in Medical Imaging

Optoacoustic tomography is a powerful hybrid bioimaging method which retains rich optical contrast and diffraction-limited ultrasonic resolution at depths of varying from millimeters to several centimeters in biological tissue irrespective of photon scattering. Optoacoustic imaging is commonly performed with high power optical pulses whose absorption leads to instantaneous temperature increase, thermal expansion and, subsequently, to the generation of a pressure field proportional to the distribution of the absorbed energy. For tomographic data acquisition, the optoacoustically generated waves are detected on a surface surrounding the imaged region. Recovery of the initially generated pressure distribution from the detected tomographic projections, and hence of the optical energy deposition in the tissue, constitutes the inverse problem of optoacoustic tomography, which is often solved using closed-form inversion formulae. However, those closed-form solutions are only exact for ideal detection geometries, which often do not accurately represent the experimental conditions. Model-based image-reconstruction techniques represent an alternative approach to solving the inverse problem that can significantly reduce image artifacts associated with approximated analytical formulae and significantly enhance image quality in non-ideal imaging scenarios. In the model-based reconstruction, a linear forward model is constructed to accurately describe the experimental conditions of the imaging setup. Inversion is performed numerically and may include regularization when the projection data is insufficient. This chapter demonstrates the benefits of the model-based reconstruction approach and describes numerically efficient methods for its implementation.
[Read More…]

Fig. 5 A comparison between back-projection and model-based reconstructions of a mouse heart in a 3D limited view scenario. The reduction of streak aritifacts in the model-based reconstruction
is readily seen in the images.

A. Rosenthal, D. Razansky, V. Ntziachristos, „Model-based image reconstruction in optoacoustic tomography,“ in Computer Vision in Medical Imaging, edited by C.H. Chen; World Scientific Publishing, October 2013.