Related Experiment Videos
Leveraging Attention-Based Deep Learning in Binary Classification for Early-Stage Breast Cancer Diagnosis
Lama A Aldakhil1, Shuaa S Alharbi1, Abdulrahman Aloraini1
1Department of Information Technology, College of Computer, Qassim University, Buraydah 52571, Saudi Arabia.
Diagnostics (Basel, Switzerland)
|March 28, 2025
Summary
Integrating attention-based deep learning with traditional machine learning improves breast cancer image classification. Lower magnification levels enhance diagnostic accuracy and model generalization for better early detection.
Area of Science:
- Medical Imaging
- Computational Pathology
- Artificial Intelligence in Medicine
Background:
- Breast cancer diagnosis presents a global challenge, necessitating advancements in early detection accuracy and efficiency.
- Histopathological image analysis is crucial for accurate breast cancer classification.
- Innovative computational methods are needed to enhance diagnostic capabilities.
Purpose of the Study:
- To investigate the integration of attention-based deep learning models with traditional machine learning (ML) for classifying histopathological breast cancer images.
- To evaluate the performance of the Efficient Channel-Spatial Attention Network (ECSAnet) in binary classification tasks.
- To analyze the impact of different magnification levels on classification accuracy, robustness, and generalization.
Main Methods:
- Utilized the BreakHis dataset comprising histopathological images of benign and malignant breast tumors at 40×, 100×, 200×, and 400× magnifications.
- Employed the Efficient Channel-Spatial Attention Network (ECSAnet) with advanced attention mechanisms for feature extraction.
- Evaluated ECSAnet independently and in conjunction with traditional ML models like Decision Trees and Logistic Regression.
Main Results:
- Lower magnification levels (40×, 100×) demonstrated superior accuracy, robustness, and generalization compared to higher magnifications for binary classification.
- Combining ECSAnet with traditional ML models yielded improved classification performance, particularly at lower magnification levels.
- Attention-based models showed significant diagnostic strengths, highlighting the importance of magnification selection.
Conclusions:
- Attention-based deep learning models like ECSAnet, when integrated with traditional ML, show significant potential for improving breast cancer diagnostics.
- Lower magnification levels are diagnostically advantageous for histopathological breast cancer image analysis.
- Future research should explore hybrid architectures and multimodal approaches to further enhance diagnostic accuracy.