Related Experiment Video
Updated: May 26, 2026
09:31
Automated Quantification of Hematopoietic Cell – Stromal Cell Interactions in Histological Images of Undecalcified Bone
Published on: April 8, 2015
Robust segmentation of overlapping cells in histopathology specimens using parallel seed detection and repulsive
Xin Qi1, Fuyong Xing, David J Foran
1Department of Pathology and Laboratory Medicine, University of Medicine and Dentistry New Jersey (UMDNJ)-Robert Wood Johnson Medical School, Piscataway, NJ 08854, USA. xinqi2000@gmail.com
IEEE Transactions on Bio-Medical Engineering
|December 15, 2011
Summary
This study introduces a new automated algorithm for segmenting touching cells in breast cancer histopathology images. The method accurately separates overlapping cells, aiding in early cancer detection and quantitative analysis.
Area of Science:
- Digital Pathology
- Computational Biology
- Medical Image Analysis
Background:
- Automated histopathology image analysis can aid breast cancer detection and characterization.
- Accurate cell segmentation in tissue microarrays (TMAs) is crucial for quantitative analysis but challenged by cell overlap.
Purpose of the Study:
- To develop a novel algorithm for reliable separation of touching cells in breast cancer TMAs.
- To enable accurate quantitative analysis for improved breast cancer diagnostics.
Main Methods:
- A two-step algorithm combining single-path voting with mean-shift clustering for cell center localization.
- A level set algorithm with an interactive model for precise cell contour segmentation.
- Implementation of a parallel GPU-accelerated version for enhanced processing speed.
Main Results:
- The algorithm successfully segmented touching cells in dense breast cancer TMA specimens.
- Pixel-wise accuracy was validated against human expert annotations.
- Significant speedup was achieved with the GPU-parallelized version compared to C/C++ implementation.
Conclusions:
- The proposed automated segmentation algorithm effectively addresses the challenge of overlapping cells in histopathology.
- This method provides a robust foundation for quantitative analysis in breast cancer research and diagnostics.
- The GPU implementation offers a computationally efficient solution for large-scale image analysis.
