Expectation-maximization technique for fibro-glandular discs detection in mammography images

Area of Science:

• Medical Imaging
• Biostatistics
• Computational Biology

Background:

• Breast cancer is a leading cause of mortality in women globally.
• Mammography is crucial for early breast cancer detection, especially for non-palpable lesions.
• Accurate data modeling is essential for developing effective mammography analysis tools.

Purpose of the Study:

• To investigate the suitability of Gamma distribution for modeling mammography image data.
• To develop a robust method for threshold estimation in mammography images.
• To enhance the accuracy of fibro-glandular disc detection in mammograms.

Main Methods:

• Utilizing Gamma distribution for modeling mammography image data, recognizing histograms as mixtures of Gamma distributions.
• Implementing threshold selection at the valleys of multi-modal histograms based on statistical parameters.
• Developing the expectation-maximization technique with gamma distribution (EMTG) for estimating statistical histogram parameters.

Main Results:

• The Gamma distribution is demonstrated to be more suitable than Gaussian distribution for mammography image data.
• The EMTG technique effectively estimates statistical histogram parameters.
• Experimental results show improved accuracy in detecting fibro-glandular discs using the proposed technique.

Conclusions:

• The EMTG technique offers a statistically sound approach for analyzing mammography image data.
• This method enhances the precision of identifying key structures like fibro-glandular discs.
• The findings contribute to advancing automated breast cancer detection systems.