Related Experiment Video
Updated: Feb 22, 2026
07:05
Area-based Image Analysis Algorithm for Quantification of Macrophage-fibroblast Cocultures
Published on: February 15, 2022
3.0K
Determining Image Processing Features Describing the Appearance of Challenging Mitotic Figures and Miscounted
Ziba Gandomkar1, Patrick C Brennan1, Claudia Mello-Thoms1,2
1Medical Image Optimisation and Perception Research Group (MIOPeG), Discipline of Medical Radiation Sciences, Faculty of Health Sciences, University of Sydney, Australia.
Journal of Pathology Informatics
|October 3, 2017
Summary
Quantitative features can differentiate challenging mitoses from easily identifiable ones and miscounted objects in breast slides. Gabor textural and intensity-based features showed the most significant differences.
Area of Science:
- Digital pathology
- Computational biology
- Breast cancer research
Background:
- Pathologist agreement on mitosis identification in breast slides is limited.
- Accurate mitosis counting is crucial for breast cancer prognostication.
Purpose of the Study:
- To identify quantitative features distinguishing easily identifiable mitoses, challenging mitoses, and miscounted non-mitotic objects in breast slides.
- To evaluate different color spaces for their ability to capture these distinctions.
Main Methods:
- Mitoses and non-mitotic objects were segmented using k-means clustering and morphological operations.
- Morphological, intensity, and textural features were extracted from segmented areas and image patches across eight color spaces.
- Statistical analysis (Kruskal-Wallis H-test, Tukey-Kramer test) identified significant feature differences.
Main Results:
- Challenging mitoses were smaller and rounder than easily identifiable ones.
- Gabor textural features best differentiated challenging from easily identifiable mitoses.
- Non-mitotic objects were similar in size to easily identifiable mitoses but rounder.
- Intensity features from chromatin channels were most discriminative between easily identifiable mitoses and miscounted objects.
Conclusions:
- Quantitative image features can characterize challenging mitoses and miscounted non-mitotic objects.
- This approach may aid in improving the accuracy and consistency of mitosis counting in breast pathology.
