Related Experiment Video
Updated: Sep 24, 2025
10:25
Deep Learning-Based Segmentation of Cryo-Electron Tomograms
Published on: November 11, 2022
9.4K
Extending U-Net Network for Improved Nuclei Instance Segmentation Accuracy in Histopathology Images.
Gani Rahmon1, Imad Eddine Toubal1, Kannappan Palaniappan1
1Dept. of Electrical Engineering and Computer Science, University of Missouri, Columbia, MO 65211, USA.
IEEE Applied Imagery Pattern Recognition Workshop : [Proceedings]. IEEE Applied Imagery Pattern Recognition Workshop
|May 4, 2022
Summary
Accurate nuclei segmentation is crucial for disease analysis. A novel USE-Net model enhances segmentation by recalibrating features and outputting shape markers, improving accuracy even with limited data.
Area of Science:
- Computational pathology
- Medical image analysis
- Deep learning
Background:
- Accurate nuclei segmentation is essential for analyzing tissue samples and predicting disease progression.
- Current methods struggle with diverse nuclear morphologies and dense clustering, leading to annotation challenges and limited data for supervised learning.
Purpose of the Study:
- To develop a robust nuclei segmentation method addressing limitations of current approaches.
- To improve the separation of densely clustered nuclei and enhance segmentation accuracy.
Main Methods:
- Utilized a U-Net architecture enhanced with Squeeze and Excitation (USE-Net) blocks for feature recalibration.
- Extended the network to output shape markers alongside segmentation masks for improved nuclei separation.
- Trained and evaluated the model on the 2018 MICCAI Multi-Organ-Nuclei-Segmentation (MoNuSeg) dataset.
Main Results:
- Achieved promising nuclei segmentation results on unseen data.
- Demonstrated robust performance despite a significantly small training dataset.
- The USE-Net model effectively separated densely clustered nuclei.
Conclusions:
- The proposed USE-Net offers a robust solution for nuclei instance segmentation in computational pathology.
- The method shows potential for accurate analysis even with limited annotated data, overcoming common challenges in the field.
