Breast Cancer Histopathological Image Classification Based on Graph Assisted Global Reasoning

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Breast cancer is the most common cancer in women globally.
• Histopathological image analysis is crucial for accurate tumor detection.
• Current methods require further enhancement for high-precision pathological diagnosis.

Purpose of the Study:

• To develop a deep learning model for automated breast cancer histopathological feature extraction and tumor identification.
• To assist clinicians in achieving high-precision pathological diagnoses.
• To improve the accuracy and efficiency of breast cancer detection through automated analysis.

Main Methods:

• A novel dual-stream global-local network (DSGLNet) was proposed for histopathological image classification.
• DSGLNet utilizes a convolutional network for local feature extraction and graph convolutional mapping for global feature interaction.
• Image preprocessing involved feature engineering for color normalization and boundary enhancement.

Main Results:

• The DSGLNet model demonstrated strong performance on the BreakHis dataset.
• At 40x magnification, the model achieved 0.966 accuracy and 0.973 precision in identifying tumor nature.
• The proposed method outperformed other advanced techniques in breast cancer histopathological image classification.

Conclusions:

• DSGLNet effectively integrates local and global features for precise image classification.
• The model shows significant potential in assisting pathologists with accurate breast cancer diagnosis.
• The study highlights the efficacy of deep learning in advancing histopathological analysis for oncology.