Shape-based multi-region segmentation framework: application to 3D infants MRI data

International audience<p>This paper presents a novel shape-guided multi-region variational region growing framework for extracting simultaneously thoracic and abdominal organs on 3D infants whole body MRI. Due to the inherent low qualityof these data, classical segmentation methods tend to fail at the multi-segmentation task. To compensate forthe low resolution and the lack of contrast and to enable the simultaneous segmentation of multiple organs, weintroduce a segmentation framework on a graph of supervoxels that combines supervoxels intensity distributionweighted by gradient vector flow value and a shape prior per tissue. The intensity-based homogeneity criteriaand the shape prior, encoded using Legendre moments, are added as energy terms in the functional to be op-timized. The intensity-based energy is computed using both local (voxel value) and global (neigboring regionsmean values, adjacent voxels values and distance to the neighboring regions) criteria. Inter-region conflict resolution is handled using a weighted Voronoi decomposition method, the weights being determined using tissuesdensities. The energy terms of the global energy equation are weighted using an information on growth directionand on gradient vector flow value in order to either guide the segmentation toward the image natural edges ifit is consistent with image and shape prior terms or enforce the shape prior term otherwise. Results on 3Dinfants MRI data are presented and compared to a set of manual segmentations. Both visual comparison andquantitative measurements show good results.