Related Experiment Video
Updated: Sep 21, 2025
09:31
Automated Quantification of Hematopoietic Cell – Stromal Cell Interactions in Histological Images of Undecalcified Bone
Published on: April 8, 2015
11.7K
Deep Features Aggregation-Based Joint Segmentation of Cytoplasm and Nuclei in White Blood Cells
IEEE Journal of Biomedical and Health Informatics
|May 31, 2022
Summary
We developed two novel AI networks, LDS-Net and LDAS-Net, to automate the segmentation of white blood cells (WBCs). These networks accurately identify cell cytoplasm and nuclei, improving upon manual microscopic analysis.
Area of Science:
- Medical Imaging
- Computational Biology
- Artificial Intelligence
Background:
- Manual microscopic inspection of white blood cells (WBCs) for diagnosis is time-consuming and prone to errors.
- Automated analysis of WBCs is crucial for efficient and accurate hematological assessments.
Purpose of the Study:
- To introduce two novel shallow neural networks, LDS-Net and LDAS-Net, for automated joint segmentation of WBC cytoplasm and nuclei.
- To improve the accuracy and efficiency of WBC image analysis compared to traditional methods.
Main Methods:
- Development of two shallow convolutional neural networks: LDS-Net and LDAS-Net.
- LDAS-Net incorporates a novel bridge for low-level feature transfer and a dense feature concatenation block.
- Evaluation on four publicly available WBC datasets.
Main Results:
- High dice coefficients achieved for cytoplasmic segmentation (up to 99.0%) and nuclei segmentation (up to 98.09%).
- The proposed method outperforms existing state-of-the-art techniques.
- The networks demonstrate superior computational efficiency with only 6.5 million trainable parameters.
Conclusions:
- LDS-Net and LDAS-Net provide an efficient and accurate automated solution for WBC cytoplasm and nuclei segmentation.
- The developed models offer a significant advancement over manual methods in hematology.
- These networks have the potential to enhance diagnostic workflows in clinical settings.
