Automatic detection and segmentation of kidneys in 3D CT images using random forests

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Nephrology

Background:

• Accurate kidney segmentation in 3D CT images is crucial for extracting clinical information.
• Existing algorithms often lack speed, automation, or robustness to variations in contrast enhancement and field of view.

Purpose of the Study:

• To develop a fast, automatic, and robust algorithm for kidney segmentation in 3D CT images.
• To combine random forests and template deformation for improved segmentation accuracy and clinical applicability.

Main Methods:

• Kidney localization using a coarse-to-fine strategy with random forests.
• Refinement of kidney positions via a cascade of local regression forests.
• Probabilistic segmentation using a classification forest, followed by implicit template deformation.

Main Results:

• The algorithm achieved accurate detection and segmentation of 80% of kidneys (Dice coefficient > 0.90).
• Segmentation was performed rapidly, in a few seconds per volume.
• Validation on a diverse dataset of 233 CT scans from 89 patients demonstrated robustness.

Conclusions:

• The combined approach of random forests and template deformation provides an effective solution for automated kidney segmentation.
• The developed algorithm meets clinical requirements for speed, automation, and robustness.
• This technique holds promise for enhancing diagnostic capabilities in nephrology.