4D segmentation algorithm with application to 3D+time image segmentation (Q2111549)

The authors propose a segmentation approach that is based on the surface evolution governed by a nonlinear partial differential equation. The approach is a generalization of the SUBSURF (Subjective Surface) model [\textit{A. Sarti} et al., Proc. Natl. Acad. Sci. USA 97, No. 12, 6258--6263 (2000; Zbl 0966.68214)]. This generalization is achieved by defining the input to the edge detector function as the weighted sum of norm of presmoothed 4D image and norm of presmoothed thresholded 4D image in a local neighborhood. A semi-implicit finite volume method is introduced and studied. The reduced diamond cell approach is used for approximating the gradient of the solution. The unconditional stability is proved. In the special case of the 3D framework, the studied 4D method is applied to 3D image segmentation of cell nuclei and membranes within the pectoral fin of developing zebrafish embryo and cell nuclei in developing mouse embryo. The OpenMP and MPI for the studied model's efficient (parallel) computer implementation are employed. Several numerical examples are presented to demonstrate how the model works both for 3D and 4D images.