Related Experiment Video
Updated: Aug 4, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
CUSS-Net: A Cascaded Unsupervised-Based Strategy and Supervised Network for Biomedical Image Diagnosis and
IEEE Journal of Biomedical and Health Informatics
|April 6, 2023
Summary
This study introduces CUSS-Net, a novel cascaded strategy for automated white blood cell (WBC) and skin lesion segmentation and classification. CUSS-Net enhances deep learning models by integrating unsupervised and supervised tasks, achieving superior performance on medical image analysis.
Area of Science:
- Medical image analysis
- Computer-aided diagnosis
- Deep learning
Background:
- Biomedical image segmentation and classification are vital for computer-aided diagnosis.
- Current deep convolutional neural networks (CNNs) often focus on single tasks, neglecting synergistic potential of multi-task learning.
- Automated analysis of white blood cell (WBC) and skin lesion images requires robust segmentation and classification methods.
Purpose of the Study:
- To propose a cascaded unsupervised-based strategy (CUSS-Net) to enhance supervised CNN frameworks for automated WBC and skin lesion segmentation and classification.
- To improve the accuracy of segmentation and classification by leveraging multi-task learning and a novel network architecture.
Main Methods:
- Developed CUSS-Net, comprising an unsupervised-based strategy (US) module, an enhanced segmentation network (E-SegNet), and a mask-guided classification network (MG-ClsNet).
- The US module generates coarse masks to guide E-SegNet for accurate object localization and segmentation.
- Employed a novel cascaded dense inception module for enhanced feature extraction and a hybrid loss function (dice loss + cross-entropy loss) to address class imbalance.
Main Results:
- CUSS-Net demonstrated superior performance compared to existing state-of-the-art methods on three public medical image datasets.
- The cascaded approach effectively integrated unsupervised pre-localization with supervised segmentation and classification.
- The enhanced coarse masks from E-SegNet significantly improved the classification accuracy in MG-ClsNet.
Conclusions:
- The proposed CUSS-Net effectively boosts supervised CNN performance for medical image segmentation and classification through a cascaded unsupervised-supervised strategy.
- CUSS-Net offers a promising approach for automated analysis of white blood cells and skin lesions in computer-aided diagnosis systems.
- The integration of multi-task learning and advanced network modules contributes to improved accuracy and robustness.
