Breast skin-line detection using dynamic programming.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Image Processing

Background:

• Accurate breast skin-line extraction is crucial for computer-aided detection (CAD) systems.
• Challenges include noise, low contrast at the skin-air interface, and artifacts.
• Existing methods struggle with these common mammogram imperfections.

Purpose of the Study:

• To develop a novel, robust method for breast skin-line extraction using dynamic programming.
• To improve the accuracy and reliability of skin-line detection in mammograms.
• To address limitations of current preprocessing techniques in CAD systems.

Main Methods:

• Utilized stroma edge information to constrain the search for the breast contour.
• Employed the Laplace operator to identify candidate border points, mitigating noise.
• Applied a dynamic programming algorithm with a cost function minimization for optimal contour detection.

Main Results:

• The proposed method successfully extracted breast skin-lines from 82 mammograms.
• Evaluation using the Polyline Distance Measure showed a mean error of 2.05 pixels.
• A standard deviation of 0.80 pixels indicates high consistency and accuracy.

Conclusions:

• The novel dynamic programming approach provides an accurate and robust method for breast skin-line extraction.
• This technique offers significant improvements for preprocessing in mammography CAD systems.
• The method demonstrates high performance in handling challenging imaging conditions.