3D cell nuclei segmentation based on gradient flow tracking.

Area of Science:

• Microscopy and Imaging Science
• Computational Biology
• Cell Biology

Background:

• Accurate cell nuclei segmentation in 3D microscopy is crucial for biological research.
• Existing methods struggle with closely juxtaposed or touching nuclei.
• A novel, fully automated segmentation method is proposed.

Purpose of the Study:

• To develop and validate a novel, automated method for segmenting cell nuclei in 3D microscopic images.
• To specifically address the challenge of segmenting closely juxtaposed or touching nuclei.
• To demonstrate the method's performance and generalizability on diverse datasets.

Main Methods:

• The method employs a three-stage approach: gradient diffusion, gradient flow tracking and grouping, and local adaptive thresholding.
• This automated pipeline processes 3D microscopic image data.
• The algorithm is designed for robustness in complex imaging scenarios.

Main Results:

• The proposed method demonstrates high accuracy, achieving over 90% volume overlap compared to expert manual segmentation.
• Both over-segmentation and under-segmentation rates are maintained around 5%.
• Qualitative and quantitative results confirm the method's performance on synthesized and original 3D images.

Conclusions:

• The developed algorithm effectively segments closely juxtaposed or touching cell nuclei from 3D microscopy images.
• The method provides reasonable accuracy for challenging segmentation tasks.
• This automated approach offers a valuable tool for biological studies relying on 3D nuclear segmentation.