Solution of the missing cone problem by artificial neural network

This paper considers the well-known limited angle or missing cone problem. We present an alternative method to restore a complete set of projection data from limited angle of views without prior information of the scanning object, using two-dimensional sampling theory and the result by Rattey and Lindgren (1981) which shows that the spectral support of CT projection data is bowtie-shaped. A general matrices formulation is developed and the problem is posed as an least square optimization problem. This problem is then implemented by a nonstandard neural network. A novel training algorithm is proposed minimizing a modified error criterion. Computer simulation results are presented to demonstrate the validity of the algorithm