Fast automated segmentation of multiple objects via spatially weighted shape learning.

摘要: Active shape models (ASMs) have proved successful in automatic segmentation by using and appearance priors a number of areas such as prostate segmentation, where accurate contouring is important treatment planning for cancer. The ASM approach however, heavily reliant on good initialisation achieving high quality. This often requires algorithms with computational complexity, three dimensional (3D) image registration. In this work, we present fast, self-initialised that simultaneously fits multiple objects hierarchically controlled spatially weighted learning. Prominent are targeted initially spatial weights progressively adjusted so the next (more difficult, less visible) object initialised series models. scheme was validated compared to multi-atlas 3D magnetic resonance (MR) images 38 cancer patients had same (mean, median, inter-rater) Dice's similarity coefficients (0.79, 0.81, 0.85), while having no registration error time 12–15 min, nearly an order magnitude faster than approach.