Fully Automated Renal Tissue Volumetry in MR Volume Data Using Prior-Shape-Based Segmentation in Subject-Specific...

Area of Science:

• Medical Imaging
• Radiology
• Computational Anatomy

Background:

• Accurate organ volumetry in magnetic resonance (MR) data is crucial for epidemiological studies and clinical practice.
• Manual segmentation is time-consuming and subject to reader variability, necessitating automated methods for large-scale studies.
• Existing automated methods lack comprehensive renal tissue segmentation, particularly for cortex and medulla.

Purpose of the Study:

• To present an automated framework for segmenting renal parenchyma, cortex, and medulla in native MR volume data.
• To introduce a novel subject-specific probability map computation strategy addressing MR intensity variability.
• To enable precise organ volumetry in large-scale studies without user interaction.

Main Methods:

• Developed an automated framework incorporating subject-specific probability maps for renal tissue types.
• Utilized 2-D and 3-D prior-shape knowledge within modular framework components.
• Employed a level set segmentation strategy for renal parenchyma and fuzzy clustering for cortex/medulla delineation within parenchyma.

Main Results:

• The novel subject-specific probability map computation significantly improved tissue probability map quality compared to existing methods.
• The framework demonstrated improved results for renal parenchyma segmentation.
• Cortex and medulla segmentation showed promising results, though direct comparison to state-of-the-art is limited due to the absence of comparable automated methods.

Conclusions:

• The automated framework provides a robust solution for renal tissue segmentation in MR imaging.
• The subject-specific probability map approach enhances accuracy and addresses intensity variations.
• This method holds significant potential for advancing large-scale epidemiological studies and clinical applications requiring precise renal volumetry.