Dynamic multiple thresholding breast boundary detection algorithm for mammograms.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Image Processing

Background:

• Automated breast boundary detection is crucial for mammogram analysis.
• Accurate boundary detection aids in the early and precise diagnosis of breast diseases.

Purpose of the Study:

• To develop and evaluate a novel dynamic multiple thresholding based breast boundary (MTBB) detection method.
• To enhance the accuracy of automated breast boundary detection in digitized mammograms.

Main Methods:

• Developed a two-stage MTBB method using dynamic thresholding and Sobel filtering on 716 mammograms.
• Established a reference standard through manual boundary tracing by an experienced radiologist.
• Evaluated accuracy using Hausdorff distance, average minimum Euclidean distance, and area overlap measure.

Main Results:

• MTBB-Final achieved 94% accuracy for Hausdorff distance < 6 pixels, compared to 68% for the gradient-based method.
• 96% of images showed average minimum Euclidean distance < 1.5 pixels with MTBB-Final.
• Area overlap measure > 0.9 was achieved for 99% of images using MTBB-Final, showing statistically significant improvement (p < 0.0001).

Conclusions:

• The MTBB approach, integrating dynamic multiple thresholding and gradient information, offers superior performance.
• This method significantly enhances automated breast boundary detection accuracy in mammograms.
• The improved accuracy supports more reliable computer-aided analysis for breast cancer screening.