Morphology-based interslice interpolation using local intensity information for segmentation

The interpolation of contours between slices in the absence of the original intensity image has been a challenging task and investigated for many years. In some applications of medical imaging, however, objects of interest are segmented manually on selected slices and the intensity image is available. The latter can be used to improve the quality of interpolated segmentations. In this paper, we present a two-step approach to accurate interslice interpolation of manual segmentations using information from both object shape and image intensity. Morphology based shape interpolation followed by the application of intensity-based neighborhood voting to adjust boundary voxels were used to integrate the two information sources. We compared our method to three existing interpolation methods for magnetic resonance images of mouse and human brain. The proposed method outperformed the three methods, having lower average error rates.