Advanced image preprocessing and context-aware spatial decomposition for enhanced breast cancer segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer diagnosis relies on medical imaging, but challenges like noise, low contrast, and resolution hinder accurate segmentation of malignant sites.
• Existing segmentation methods struggle with image artifacts, impacting diagnostic reliability.

Purpose of the Study:

• To develop and evaluate an integrated solution combining Advanced Image Preprocessing Techniques (AIPT) and a Context-Aware Spatial Decomposition Network (CASDN) for improved breast cancer segmentation.
• To enhance the clarity and reduce distortions in medical images for better tumor identification.

Main Methods:

• A comprehensive AIPT pipeline was employed, including Adaptive Thresholding, Hierarchical Contrast Normalization, Contextual Feature Augmentation, Multi-Scale Region Enhancement, and Dynamic Histogram Equalization.
• The preprocessed images were then fed into the CASDN for segmentation.
• Convolutional Neural Networks were utilized for classification tasks.

Main Results:

• The proposed method achieved a Dice Coefficient of 0.89, IoU of 0.85, and Hausdorff Distance of 5.2, demonstrating superior tumor margin segmentation.
• Classification models using the enhanced preprocessing pipeline achieved 85.3% accuracy and an AUC-ROC of 0.90.
• The system showed robust compatibility across various imaging modalities like mammograms, ultrasounds, and MRI scans.

Conclusions:

• The integration of AIPT and CASDN significantly improves breast cancer segmentation accuracy and classification performance.
• The advanced preprocessing pipeline effectively mitigates image noise and distortions, leading to clearer medical images.
• The developed technique offers a robust and effective solution for breast cancer diagnosis across multiple imaging modalities.