Fast Cell Segmentation Using Scalable Sparse Manifold Learning and Affine Transform-approximated Active Contour.

Area of Science:

• Digital Pathology
• Computational Biology
• Medical Image Analysis

Background:

• Accurate cell segmentation in whole slide images of neuroendocrine tumors (NETs) is challenging due to numerous cells and indistinct boundaries.
• Existing methods struggle with efficiency and scalability for large pathology specimens.

Purpose of the Study:

• To develop a fast, high-throughput cell segmentation algorithm for NETs in whole slide images.
• To address challenges posed by cell density and weak boundaries in pathological images.

Main Methods:

• A hybrid approach combining top-down shape models and bottom-up image appearance.
• Utilizing scalable sparse manifold learning to model diverse cell shape priors.
• Employing an affine transform-approximated active contour model for efficient contour deformation.

Main Results:

• The proposed algorithm achieves high throughput and superior performance compared to state-of-the-art methods.
• Demonstrated effectiveness in segmenting neuroendocrine tumor cells in 12 whole slide images.
• Significant reduction in computational time due to the novel active contour model.

Conclusions:

• This work presents the first high-throughput cell segmentation algorithm for pathology using manifold learning to accelerate active contour models.
• The developed method offers an efficient and effective solution for NET cell segmentation in digital pathology.