Bladder Wall Segmentation and Characterization on MR Images: Computer-Aided Spina Bifida Diagnosis

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Machine Learning in Medicine

Background:

• Accurate segmentation of bladder boundaries in Magnetic Resonance Images (MRI) is essential for diagnosing bladder conditions and assessing function.
• Current methods require robust segmentation and classification techniques for reliable bladder wall analysis.
• This research focuses on developing an optimized system for bladder wall segmentation and classification using MRI data.

Purpose of the Study:

• To propose an optimized system for segmenting and classifying the bladder wall from MRI images.
• To evaluate the performance of different machine learning algorithms for bladder wall characterization.
• To enhance diagnostic capabilities for radiologists, particularly in studies related to spina bifida.

Main Methods:

• Utilized LevelSet contour-based segmentation to extract the bladder wall region of interest from T2 MRI images.
• Computed various features from the segmented bladder wall and employed automatic feature selection.
• Tested two supervised learning algorithms, Support Vector Machine (SVM) and Random Forest, optimized with a bio-inspired algorithm (Gray Wolf Optimizer - GWO).

Main Results:

• The enhanced LevelSet algorithm demonstrated significant efficiency in segmenting the bladder wall.
• The SVM classifier, optimized by GWO with a Radial Basis Function (RBF) kernel, outperformed the Random Forest algorithm in classification accuracy.
• The optimized system effectively characterizes the bladder wall using selected discriminant features.

Conclusions:

• An optimized computer-aided system for bladder wall segmentation and characterization on MRI is presented.
• The proposed system, integrating advanced segmentation and machine learning, offers a valuable tool for radiologists.
• This approach shows potential for improving diagnostic accuracy in conditions like spina bifida.