Numerical solution of the identification problem for the attenuated Radon transform

From MaRDI portal





The attenuated Radon transform \[ R\mu f(\varphi,s)\equiv \int_{x\omega= s} \exp\Biggl(- \int^\infty_0 \mu(x+ t\omega^\perp) dt\Biggr) f(x) dx= g(\varphi,s) \] of a source function \(f(x)\), serves a mathematical tool for single-photon emission computerized tomography SPECT. The exponential weighting factor in this equation describes the process of linear photon attenuation, where \(\mu(x)\) is the attenuation coefficient in the medium, which plays the role of a parameter of the transform. The identification problem for the attenuated Radon transform is to find function \(\mu\) from \(g(\varphi,s)\) with unknown \(f(x)\). In this paper a new method for the solution of the identification problem for the attenuated Radon transform is considered. The approach is based on a matrix representation of the transform. The author considers a matrix representation of the transform and formulates the corresponding discrete consistency conditions in the form of the orthogonal projection of the data vector onto the orthogonal complement of the column space of the matrix. The singular value decomposition is applied to compute the orthogonal projector and its Fréchet derivative. The numerical algorithm suggested is based on the Newton method with Tikhonov regularization. To study the performance of the approach, a numerical simulation of a cardiac SPECT imaging experiment is also discussed. The experiment with an anthropomorphic torso phantom demonstrates promising results in the reconstructon of the thorax cross section characterized by relatively large variations of the attenuation map.











This page was built for publication: Numerical solution of the identification problem for the attenuated Radon transform

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4719156)