Maximum likelihood reconstruction for tomosynthesis

We present a maximum likelihood(ML) algorithm for tomosynthesis that reconstructs a 3D radiation attenuation coefficient map from a set of radiograms taken from various view angles. The ML algorithm is developed based on the Poisson model of two dimensional radiography. The ML reconstruction yields better image quality with fewer artifacts than the wildly used filtered backprojection (FBP) reconstruction. Finally, we show representative results on simulated data.