Fast Tomographic Reconstruction from Highly Limited Data Using Artificial Neural Networks

Obtaining accurate reconstructions from a small number of projections is important in many tomographic applications. Current advanced reconstruction methods are able to produce accurate reconstructions in some cases, but they are usually computationally expensive. Here, we present a reconstruction method based on artificial neural networks, which can be viewed as a combination of fast filtered backprojection reconstructions. Since the method learns characteristics of scanned objects during the training phase, it is able to reconstruct images accurately from limited data. Results from experimental μCT data show that the new method is able to produce more accurate reconstructions than both regular filtered backprojection and the slower iterative SIRT method, while having a relatively low computational cost.