Segmentor

Segmentor enables manual refinement of three-dimensional microscopy annotations of nuclei in tissue-cleared images acquired by light sheet microscopy to produce high-quality annotations for quantitative 3D analysis and training deep learning-based nuclear segmentation algorithms.

Key Features:

• Hybrid 2D-3D Visualization: Simultaneous two-dimensional and three-dimensional visualization for editing and inspection of segmentations.
• Automatic Region Splitting: Automatic region splitting functionality tailored for three-dimensional segmentation of nuclei to separate clustered objects.
• Manual 3D Annotation Refinement: Manual refinement of object boundaries within 3D microscopy images to improve nuclear segmentation accuracy.
• Training Dataset Generation: Produces high-quality annotated datasets suitable for training deep learning-based nuclear segmentation algorithms.

Scientific Applications:

• Quantitative 3D Microscopy Analysis: Facilitates quantitative analysis of three-dimensional microscopy images through precise nuclear segmentations.
• Nucleus Annotation in Tissue-Cleared Light Sheet Data: Applied to annotation of nuclei in tissue-cleared images acquired by light sheet microscopy, including whole-mouse-brain datasets.
• Deep Learning Model Training: Generates annotated training data for development and evaluation of deep learning-based nuclear segmentation algorithms.
• Neuroscience Structural Studies: Supports investigation of structural changes in specimens and their correlation with behavioral or cognitive differences in neuroscience research.

Methodology:

Hybrid two-dimensional and three-dimensional visualization combined with automatic region splitting for 3D nuclear segmentation and interactive boundary refinement to produce annotated datasets for deep learning-based nuclear segmentation algorithms.