Optimization of Deep Learning Network Parameters Using Uniform Experimental Design for Breast Cancer...

Area of Science:

• Medical image analysis
• Computational pathology
• Artificial intelligence in oncology

Background:

• Breast cancer diagnosis relies heavily on histopathological image analysis.
• Convolutional neural networks (CNNs) show promise for automated breast cancer classification from histopathology.
• Challenges include complex CNN parameter tuning and time-intensive data processing.

Purpose of the Study:

• To simplify and enhance CNN-based breast cancer classification from histopathological images.
• To optimize CNN parameters using a uniform experimental design (UED).
• To improve classification accuracy and efficiency in breast cancer diagnosis.

Main Methods:

• Implementation of a uniform experimental design (UED) for CNN parameter optimization.
• Application of regression analysis within UED to fine-tune model parameters.
• Classification of breast cancer histopathological images using the optimized CNN model.

Main Results:

• The proposed method achieved a classification accuracy rate of 84.41%.
• UED-based parameter optimization effectively improved classification performance.
• The results demonstrated superior performance compared to similar existing methods.

Conclusions:

• The UED approach offers an effective strategy for optimizing CNN parameters in medical image analysis.
• This method enhances classification accuracy for breast cancer diagnosis.
• The findings suggest a more efficient and accurate approach to computer-aided breast cancer detection.