Segmentation of nuclei (Waahlby algorithm)

Segmentation of nuclei (Waahlby algorithm) segments nuclei in fluorescence images stained with DNA-specific dyes (e.g., DAPI) using a region-based approach that integrates morphological filtering and gradient-magnitude analysis to produce object masks for quantitative nuclear analysis.

Key Features:

• Region-based segmentation: Applies a region-based approach combining morphological filtering on the original image and its gradient magnitude.
• Seed generation: Generates object and background seeds via morphological filters to serve as initialization points for segmentation.
• Watershed segmentation: Performs watershed segmentation from the generated seeds using the image gradient magnitude.
• Refinement by merging: Merges over-segmented regions based on gradient magnitude along shared boundaries and removes poorly contrasted objects.
• Cluster separation: Separates clusters of nuclei using shape characteristics to improve object delineation.
• Minimal parameterization: Operates with a minimal set of five input parameters that can be optimized on a test image for consistent application.

Scientific Applications:

• Nucleus counting: Automated counting of nuclei in two-dimensional and three-dimensional image fields with reported validation against manual counts (~90% correct segmentation accuracy).
• Quantitative nuclear analysis: Supports cellular biology, histopathology, and other studies requiring detailed nuclear segmentation and morphometric measurements.

Methodology:

Generate object and background seeds via morphological filters, compute the gradient magnitude, run watershed segmentation from the seeds, merge regions based on gradient magnitude along boundaries, separate clusters by shape characteristics, and adjust five input parameters on a test image.