WAVELET-BASED AUTOMATIC PECTORAL MUSCLE SEGMENTATION FOR MAMMOGRAMS.

Area of Science:

• Medical Imaging
• Image Processing
• Computational Radiology

Background:

• Mammogram interpretation requires accurate breast tissue segmentation, a manual process that is time-consuming and subjective.
• Automated segmentation methods are crucial for efficient and consistent pre-processing in radiological analysis.

Approach:

• The study adapts the 2D Wavelet Transform Modulus Maxima (WTMM) method for automated breast tissue segmentation.
• This approach utilizes local maxima of the continuous Gaussian wavelet transform to identify edge detection lines (maxima chains).
• Maxima chains from multiple wavelet scales are sorted to generate a precise breast tissue segmentation mask.

Key Points:

• The 2D WTMM method achieved a high median Dice-Sorenson Coefficient (DSC) of 0.9763 on 1042 mammographic views.
• The WTMM approach demonstrated statistically significant superiority (p=0.0067) compared to the open-source software OpenBreast.
• The method reliably delineates pectoral muscle and provides accurate whole breast tissue segmentation.

Conclusions:

• The adapted 2D WTMM segmentation method offers a reliable and accurate automated solution for breast tissue segmentation in mammography.
• This technique can enhance the efficiency and consistency of pre-processing steps in mammogram interpretation.
• The findings suggest the WTMM method as a valuable tool for computer-aided diagnosis in breast imaging.