Heart region extraction and segmentation from chest CT images using Hopfield Artificial Neural Networks

A system for extracting and segmenting heart regions from three-dimensional (3D) CT chest images is proposed in this paper. At first, the regions of interest (ROIs) are extracted using pure basic image processing techniques applied on the 2D CT slices. Secondly, the ROIs in each slice are segmented using Hopfield Artificial Neural Networks (HANN). The segmentation results include tissues belonging to the heart and its surrounding organs. To distinguish between heart regions and the non-heart regions, a rule-based filtering approach is adopted. The system is evaluated using a database of 735 chest CT slices from 5 patients. It shows a good and accurate performance with some exceptions.