Kidney segmentation from DCE-MRI converging level set methods, fuzzy clustering and Markov random field modeling

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Accurate kidney segmentation in Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) is crucial for early diagnosis of transplanted kidney function.
• Existing segmentation methods may lack accuracy or robustness, particularly with complex image characteristics.

Purpose of the Study:

• To develop an automated and accurate DCE-MRI kidney segmentation method.
• To integrate fuzzy c-means (FCM) clustering and Markov random field (MRF) modeling within a level set (LS) framework.

Main Methods:

• Proposed a novel automated kidney segmentation method for DCE-MRI.
• Integrated fuzzy c-means (FCM) clustering and second-order Markov random field (MRF) modeling into a level set (LS) formulation.
• Utilized fuzzy memberships, kidney shape prior, and spatial interactions to guide contour evolution.

Main Results:

• Achieved high and consistent segmentation accuracy (Dice Similarity Coefficient: 0.956 ± 0.019).
• Demonstrated superior performance in Hausdorff distance (HD95: 1.15 ± 1.46) compared to other LS methods and U-Net based deep learning models.
• Showed significant accuracy improvements on low-contrast and high-noise DCE-MRI images.

Conclusions:

• The proposed automated DCE-MRI kidney segmentation method is accurate and robust.
• This method offers a significant advancement for the preliminary analysis of transplanted kidney function.
• The technique shows particular promise for improving segmentation in challenging medical imaging scenarios.