A strategy for Compton scatter correction in brain SPECT by orthogonal image expansions assisted by neural networks

The simultaneous dual energy acquisition is often required to accurate co-registration of brain images of both energy windows. However, the reconstructed images in the lower energy window are usually distorted by a significant Compton scatter component originating from the higher energy photons. We are proposing two methods for calculating the Compton scatter correction. Both of them utilize sets of TC99m and associated Compton scatter images in TI(201) energy window that represent a patient population and calculate the sought Compton image estimate 'by comparison' of the given new Tc99m image with the Tc99m set. The first method is based on the principal component image expansion and the second one utilizes in addition learning neural networks. Means to measure the accuracy of results are proposed as well. The proposed methods can be modified for application to other anatomical domains.