Related Experiment Video
Updated: Jun 24, 2026
06:27
Automated Analysis of C. elegans Fluorescence Images using SegElegans
Published on: October 10, 2025
Quantitative Detection of Entotic Cell-In-Cell Structures Using Deformable Segmentation and Deep Learning
Maria V Leyba-Mesa1, Mikołaj Biegański2, Elijah Ray1
1Department of Biomedical Engineering, College of Engineering and Polymer Science, The University of Akron, Akron, Ohio, USA.
Summary
Automated cell-in-cell (CIC) detection using computational methods is crucial for cancer research. This study developed deformable segmentation and YOLOv8 models, showing deformable models excel in sensitivity for entosis identification.
Area of Science:
- Cell Biology
- Computational Pathology
- Cancer Research
Background:
- Cell-in-cell (CIC) structures, especially entosis, are increasingly linked to cancer progression and patient outcomes.
- Manual identification of entosis is labor-intensive, hindering systematic investigation.
- Scalable and reproducible computational methods are needed for entosis detection.
Purpose of the Study:
- To develop and evaluate computational models for automated entotic cell identification.
- To compare the performance of morphology-driven deformable segmentation models with a deep learning-based YOLOv8 framework.
Main Methods:
- Developed five morphology-driven deformable segmentation models capturing features like entropy, spatial proximity, contour topology, and circularity.
- Implemented a YOLOv8 deep learning framework for entotic cell detection.
- Evaluated models on BxPC3 (pancreatic) and MCF7 (breast) cancer cell lines.
Main Results:
- Deformable Models A and C demonstrated high sensitivity (0.91-0.94 recall, 0.81-0.83 F1-score) for entosis detection.
- YOLOv8 achieved high accuracy (0.97) and specificity (0.98) but lower recall (0.65) and F1-score (0.59) due to class imbalance.
- Deformable models offered detailed segmentation, while YOLOv8 provided computational efficiency.
Conclusions:
- Morphology-driven deformable models and YOLOv8 possess complementary strengths for entosis analysis.
- Hybrid frameworks combining these approaches can enable scalable and automated entosis detection.
- Further development of computational tools is essential for advancing entosis research in cancer.
