Square-cut: a segmentation algorithm on the basis of a rectangle shape.

Area of Science:

• Medical Imaging
• Computer Vision
• Image Segmentation

Background:

• Traditional graph-based segmentation algorithms often use uniform node distribution, limiting their ability to accurately segment objects with indistinguishable boundaries from the background.
• Manual segmentation of medical images, such as vertebrae in MRI scans, is time-consuming and requires expert clinical knowledge.

Purpose of the Study:

• To develop and evaluate a novel rectangle-based segmentation algorithm for improved object segmentation in medical imaging.
• To address the limitations of uniform node distribution in graph cuts for complex segmentation tasks.
• To provide an automated solution for vertebrae segmentation in MRI datasets, reducing the burden on physicians.

Main Methods:

• A rectangle-based segmentation approach utilizing graph cuts was developed.
• Graph nodes were sampled non-uniformly and non-equidistantly, guided by the object's rectangular shape.
• The algorithm was evaluated on Magnetic Resonance Imaging (MRI) datasets focusing on vertebrae segmentation.

Main Results:

• The proposed algorithm demonstrated superior performance in segmenting objects with areas indistinguishable from the background.
• Achieved an average Dice Similarity Coefficient (DSC) of 90.97±2.2% for vertebrae segmentation in MRI.
• The non-uniform node sampling strategy effectively improved segmentation accuracy compared to uniform methods.

Conclusions:

• The rectangle-based segmentation algorithm offers a significant advancement in automated medical image segmentation.
• This method provides a more accurate and efficient alternative to manual segmentation for clinical applications.
• The non-uniform node distribution strategy is crucial for enhancing segmentation performance in challenging imaging scenarios.