Urinary bladder segmentation in CT urography (CTU) using CLASS.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Urology

Background:

• Bladder cancer diagnosis relies on accurate imaging.
• Segmentation of bladder on CT urography (CTU) is challenging due to contrast variations.
• Computer-aided diagnosis (CAD) systems require precise anatomical segmentation.

Purpose of the Study:

• To develop and evaluate a computerized system for bladder segmentation on CTU.
• To improve computer-aided diagnosis of bladder cancer through accurate segmentation.
• To address the challenges of segmenting bladder regions with and without contrast.

Main Methods:

• A Conjoint Level set Analysis and Segmentation System (CLASS) was developed.
• CLASS employs preprocessing, 3D/2D level set segmentation, and postprocessing.
• Separate segmentation of contrast (C) and non-contrast (NC) regions, followed by contour union.

Main Results:

• CLASS achieved an average quality rating of 6.5 ± 1.3 across 81 CTU scans.
• 39 bladders received quality ratings of 7 or higher.
• CLASS demonstrated superior performance compared to a single level set method in key segmentation metrics.

Conclusions:

• The Conjoint Level set Analysis and Segmentation System (CLASS) shows significant potential for bladder segmentation in CTU.
• The developed system can aid in the computer-aided diagnosis of bladder cancer.
• CLASS effectively handles variations in bladder contrast for improved segmentation accuracy.