Reconstrução de Imagens Tomográficas Geradas por Projeções Fan-Beam com Detetores Eqüidistantes usando Redes Neurais Parcialmente Conectadas

Abstract

We present a neural network approach for the tomographic imaging problem using interpolation methods and fan-beam projections. This approach uses a partially connected neural network especially assembled for solving the tomographic reconstruction problem (without the need for training). We extended the calculations to perform the reconstruction with interpolation and to allow reconstruction of fan-beam geometry problems. The main goal is to aggregate speed and quality for the tomographic reconstruction process.