Enhancement Technique Based on the Breast Density Level for Mammogram for Computer-Aided Diagnosis

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Mammogram mass detection is challenged by overlapping dense fibroglandular tissue.
• Varying breast densities can hinder feature extraction and reduce AI model accuracy.

Purpose of the Study:

• To propose a novel textural-based image enhancement technique, Spatial-based Breast Density Enhancement for Mass Detection (SbBDEM).
• To improve the detection, segmentation, and classification of breast masses, particularly in dense breast regions.

Main Methods:

• SbBDEM optimizes image contrast and texture features based on breast density levels.
• It utilizes the Blind/Reference-less Image Spatial Quality Evaluator (BRISQUE) for parameter optimization.
• A modified You Only Look Once v3 (YOLOv3) architecture with enhanced anchor boxes is used for mass detection.

Main Results:

• SbBDEM improved mean Average Precision (mAP) by 17.24% compared to non-enhanced images.
• Achieved 94.41% accuracy for mass segmentation.
• Reached 96% accuracy for benign and malignant mass classification.

Conclusions:

• Enhancing mammograms based on breast density significantly boosts AI system performance.
• The SbBDEM technique shows potential to aid in improved clinical diagnosis of breast cancer.