Computed tomography based on a self-organizing neural network

We propose a method based on a self-organizing neural network (SONN) for computed tomography (CT). An expectation maximization-maximum likelihood algorithm, which is a well-known method for CT, is used as a learning algorithm of the network. The network is trained to minimize the Euclidean distance between the obtained projections and the projections of the estimate. Since the SONN starts with many different estimates, it is easy to obtain a global optimum.