A Dynamic Graph Cuts Method with Integrated Multiple Feature Maps for Segmenting Kidneys in 2D Ultrasound Images

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Kidney segmentation in ultrasound (US) images is challenging due to noise, low contrast, and variations.
• Texture features can enhance segmentation performance in US imaging.

Purpose of the Study:

• To develop a novel graph cuts method for accurate kidney segmentation in US images.
• To integrate image intensity and texture features for improved segmentation robustness.

Main Methods:

• A new graph cuts-based method was developed, integrating Gabor filter-extracted texture maps with image intensity.
• A localized graph of pixels near the kidney boundary was constructed for computational efficiency.
• Localized regional information was used to enhance robustness to weak boundaries, with iterative convergence.

Main Results:

• The proposed method achieved high segmentation performance for bilateral kidneys (Dice index=0.9446, accuracy=0.9919).
• Results were significantly better than existing methods (P<.05).

Conclusions:

• The novel graph cuts method demonstrates promising performance for kidney segmentation in 2D US images.
• This technique facilitates kidney characteristic extraction for predicting clinical outcomes, such as chronic kidney disease progression.