Renal compartment segmentation in DCE-MRI images.

Area of Science:

• Medical Imaging
• Radiology
• Biomedical Engineering

Background:

• Accurate segmentation of internal renal structures from DCE-MRI is crucial for functional kidney evaluation.
• Existing methods often focus on whole-kidney segmentation or CT images, lacking automated solutions for DCE-MRI internal compartment segmentation.

Purpose of the Study:

• To develop an effective and automatic method for segmenting renal cortex, medulla, and renal pelvis from DCE-MRI images.
• To achieve high segmentation accuracy with minimal manual operations and parameter settings.

Main Methods:

• Image preprocessing to reduce motion artifacts and enhance kidney regions.
• Kidney segmentation using Maximally Stable Temporal Volume (MSTV) for robustness to noise and shape variations.
• Voxel analysis using principal components and k-means clustering for compartment separation, followed by automated labeling and iterative refinement.

Main Results:

• The proposed method achieved high segmentation accuracy for internal renal structures across diverse DCE-MRI data.
• Results closely matched manual segmentations and outperformed five existing methods in experiments.
• The method demonstrated robustness to noise and adaptability to variations in kidney shape.

Conclusions:

• The developed method offers a robust, accurate, and automated solution for renal compartment segmentation in DCE-MRI.
• This advancement facilitates improved functional kidney evaluation through precise internal structure delineation.
• The technique requires minimal user intervention, making it practical for clinical application.