Hierarchical, model-based merging of multiple fragments for improved three-dimensional segmentation of nuclei.

摘要: Background Automated segmentation of fluorescently labeled cell nuclei in three-dimensional confocal images is essential for numerous studies, e.g., spatiotemporal fluorescence situ hybridization quantification immediate early gene transcription. High accuracy and automation levels are required high-throughput large-scale studies. Common sources error include tight clustering fragmentation nuclei. Previous region-based methods limited because they perform merging two nuclear fragments at a time. To achieve higher without sacrificing scale, more sophisticated yet computationally efficient algorithms needed. Methods A recursive tree-based algorithm that can consider multiple object simultaneously described. Starting with oversegmented data, it searches efficiently the optimal pattern guided by quantitative scoring criterion based on modeling. Computation bounded limiting depth tree. Results The proposed method was found to consistently better, achieving range 92% 100% compared our previous algorithm, which varied 75% 97%, even modest tree 3. The overall average improved from 90% 96%, roughly same computational cost set representative drawn CA1, CA3, parietal cortex regions rat hippocampus. Conclusion Hierarchical model-based significantly improve automated speed. © 2004 Wiley-Liss, Inc.