Xception Convolutional Deep Maxout Network for Enhanced Breast Cancer Classification Using Histopathological Images.

Area of Science:

• Oncology
• Computer Science
• Medical Imaging

Background:

• Breast cancer is a leading global cancer, necessitating accurate detection and classification for effective treatment.
• Histopathological imaging is crucial for breast cancer diagnosis, but current deep learning methods struggle with high-resolution images and subtle variations.
• Overfitting and feature mining difficulties hinder the performance of existing deep learning schemes in breast cancer classification.

Purpose of the Study:

• To develop an advanced deep learning model, the Xception Convolutional Deep Maxout Network (Xcov-DMN), for precise breast cancer classification using histopathological images.
• To address the limitations of existing deep learning models, including overfitting and challenges in extracting key features from complex medical images.

Main Methods:

• The proposed Xcov-DMN integrates the Xception Convolutional Neural Network (XCovNet), Deep Maxout Network (DMN), and Fractional Calculus (FC).
• Preprocessing involved Mean-Shift Filter for image enhancement and White Blood Cell Network (WBC-Net) with Balanced Cross-Entropy (BCE) and Focal Loss for accurate blood cell segmentation.
• Feature extraction included Colored Histograms, shape features, Haralick Texture Features, and Complete Local Binary Pattern (CLBP).

Main Results:

• The Xcov-DMN model achieved a highest accuracy of 92.755% with 90% learning data.
• The model demonstrated a True Negative Rate (TNR) of 91.977% and a True Positive Rate (TPR) of 94.765%.
• These results indicate superior performance in classifying breast cancer compared to existing deep learning approaches.

Conclusions:

• The developed Xcov-DMN model shows significant promise for accurate and reliable breast cancer classification from histopathological images.
• This deep learning approach effectively mines critical features from high-resolution images, overcoming limitations of prior methods.
• The high accuracy, TNR, and TPR suggest Xcov-DMN's potential to enhance early detection and improve patient treatment strategies.