種々の逐次近似型画像再構成法 Various Methods of Iterative Least-Squares Image Reconstruction for Emission Tomography
This paper intends to explain iterative image reconstruction methods for emission tomography based on least-squares objective functions. The methods estimate object distributions by minimizing the objective functions in iterative manners under constraints or no constraint. As for the constraints, non-negativity of pixel values is considered, as in most cases. For convenience the methods are divided into three groups. The first is the so called nonlinear optimization methods performing line searches, such as the conjugate gradient method for unconstrained optimization, and the gradient and augmented Lagrangian methods for constrained optimization. The second is the methods using the Gauss-Siedel and successive over-relaxation methods. The third is the method of applying Karush-Kuhn-Tucker conditions to image reconstruction. Algorithms for the three groups are described.
放射線医学物理 19(3), 193-204, 1999-09-30
Japan Society of Medical Physics