Related Experiment Video
Updated: May 4, 2026
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
3.9K
Breast ultrasound images for segmentation and classification using multi-task U-Net
Maddhi Anitha1,2, Ch Rajendra Prasad1, Joseph Bamidele Awotunde3,4
1Department of ECE, SR University, Warangal, Telangana, 506371, India.
Scientific Reports
|May 2, 2026
Summary
A new Multi-Task U-Net framework improves automated breast cancer detection in ultrasound images by jointly segmenting lesions and classifying tumors. This deep learning approach enhances accuracy and reliability for clinical applications.
Area of Science:
- Medical imaging
- Artificial intelligence in healthcare
- Oncology
Background:
- Breast ultrasound is crucial for early breast cancer detection, especially in dense tissues.
- Diagnostic performance is limited by operator dependency, noise, and data variability.
- Deep learning shows potential but faces challenges with small, imbalanced datasets and inconsistent annotations.
Purpose of the Study:
- To develop an integrated deep learning framework for automated breast cancer lesion segmentation and tumor classification.
- To address limitations of existing methods, including class imbalance and annotation inconsistencies.
- To improve the clinical applicability of AI in breast ultrasound diagnostics.
Main Methods:
- Proposed a Multi-Task U-Net framework for joint segmentation and classification.
- Implemented deterministic oversampling for class imbalance and a prediction-refinement module.
- Utilized attention-guided feature learning and a curated BUSI dataset for evaluation.
Main Results:
- Achieved a Dice score of 0.81 for lesion segmentation.
- Reached classification accuracy of 0.96-0.98.
- Demonstrated superior performance compared to baseline methods with good generalization.
Conclusions:
- The Multi-Task U-Net offers an effective and reliable solution for automated breast cancer detection in ultrasound.
- The framework shows strong potential for clinical integration and application.
- Joint learning strategies and data curation enhance AI model performance in medical imaging.
