A New Markov Random Field Segmentation Method for Breast Lesion Segmentation in MR images

Area of Science:

• Medical Imaging
• Computational Biology
• Radiology

Background:

• Breast cancer poses a significant global health challenge.
• Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is crucial for breast cancer management.
• Image segmentation in DCE-MRI is hindered by intensity inhomogeneities, complicating accurate analysis.

Purpose of the Study:

• To introduce an Improved-Markov Random Field (I-MRF) method for enhanced breast lesion segmentation in MR images.
• To address the limitations of conventional Markov Random Field (MRF) methods, specifically computational complexity and parameter sensitivity.

Main Methods:

• Developed an I-MRF technique that avoids iterative class membership estimation methods like Iterative Conditional Mode (ICM) and Simulated Annealing (SA).
• Modeled prior class membership distribution using a ratio of conditional probabilities for similar and non-similar pixels within a defined neighborhood.
• Implemented a non-iterative approach for maximizing posterior probability to resolve computational and sensitivity issues.

Main Results:

• The I-MRF method demonstrated superior segmentation performance compared to conventional MRF.
• Evaluations indicated improvements in accuracy and precision.
• Significant reduction in computational complexity was observed.

Conclusions:

• The proposed I-MRF method offers a more efficient and accurate solution for breast lesion segmentation in DCE-MRI.
• This technique effectively overcomes the limitations of traditional MRF-based segmentation, paving the way for improved diagnostic capabilities.