Related Experiment Video
Updated: Jan 9, 2026
06:25
Using Computer Vision Libraries to Streamline Nuclei Quantification
Published on: June 6, 2025
603
TexSegNet: An Attention-Guided Feedback-Driven Texture-Aware Deep Learning Model for Nuclei Segmentation and
Summary
TexSegNet improves nuclei detection and classification in digital pathology images, enhancing breast cancer diagnosis. This new model offers higher accuracy and better performance across diverse cell types compared to existing methods.
Area of Science:
- Computational pathology
- Digital histopathology
- Medical image analysis
Background:
- Accurate nuclei segmentation and classification are vital for breast cancer diagnosis.
- Current methods face challenges with false positives/negatives, texture representation, and segmenting clustered nuclei.
Purpose of the Study:
- To introduce TexSegNet, a novel hybrid model for nuclei instance segmentation and classification.
- To address limitations in existing computational pathology tools for breast cancer analysis.
Main Methods:
- Developed TexSegNet, a hybrid encoder-decoder model incorporating multi-scale convolutions, texture extraction, attention mechanisms, and a feedback classification branch.
- Trained on the PanNuke dataset and fine-tuned on its breast subset.
Main Results:
- TexSegNet achieved over 4% higher accuracy in nuclei detection and classification on the breast test set compared to CellViT.
- Demonstrated strong F1-scores across various cell types: neoplastic (89.3%), epithelial (91.1%), inflammatory (88.9%), and connective (84.3%).
- Maintained performance on underrepresented cell types.
Conclusions:
- TexSegNet shows significant potential as a reliable tool for digital pathology research.
- The model can serve as a decision-support tool to improve diagnostic accuracy in breast histopathology.
