Related Experiment Video
Updated: Jun 13, 2025
08:40
Quantitation of Protein Expression and Co-localization Using Multiplexed Immuno-histochemical Staining and Multispectral Imaging
Published on: April 8, 2016
12.8K
A robust image segmentation and synthesis pipeline for histopathology
Muhammad Jehanzaib1, Yasin Almalioglu2, Kutsev Bengisu Ozyoruk3
1Department of Computer Engineering, Bogazici University, Istanbul, Turkey; Department of Computer Science, FAST-NUCES, Lahore, Pakistan.
Medical Image Analysis
|September 12, 2024
Summary
PathoSeg, a novel deep learning model, enhances cancer cell segmentation in digital pathology images. It outperforms existing methods by utilizing synthetic data and attention mechanisms, improving diagnostic accuracy.
Area of Science:
- Digital Pathology
- Medical Image Analysis
- Computational Biology
Background:
- Pathology diagnostics face variability despite digital imaging advancements.
- Accurate segmentation of cancerous regions is crucial for diagnosis and treatment planning.
Purpose of the Study:
- To evaluate the performance of the PathoSeg model for automated segmentation of cancerous tissues.
- To improve diagnostic accuracy and efficiency in digital pathology.
Main Methods:
- Developed PathoSeg, a model with a modified HRNet encoder and UNet++ decoder, incorporating a CBAM attention block.
- Utilized synthetic data generated by PathopixGAN to address data imbalance.
- Created and released a new dataset with segmented masks for breast carcinoma tubules, liver steatosis, and prostate carcinoma glands.
Main Results:
- PathoSeg demonstrated superior quantitative and qualitative performance in instance and semantic segmentation compared to state-of-the-art methods.
- The use of synthetic data significantly improved PathoSeg's segmentation accuracy.
- The in-house dataset provides valuable resources for further research in histopathology image analysis.
Conclusions:
- PathoSeg offers a robust solution for accurate and automated cancer cell and tissue segmentation in digital pathology.
- The integration of synthetic data generation and attention mechanisms enhances model performance.
- The publicly released dataset and code will foster reproducibility and advance research in the field.
