Automatic Sternum Segmentation in Thoracic MRI.

Area of Science:

• Anatomy
• Medical Imaging
• Computer Vision

Background:

• The sternum, a key bone in the thoracic cage, is visible in Magnetic Resonance Imaging (MRI).
• Accurate sternum detection is crucial for 3D model reconstruction, surgical planning, and anatomical atlas development.
• Existing automated segmentation methods are limited for sternum identification in MRI.

Purpose of the Study:

• To develop and evaluate two distinct automated methods for sternum segmentation in MRI.
• To compare the efficacy of a clustering-based approach (K-Means) against a gradient-based method.

Main Methods:

• Implemented a K-Means clustering algorithm for sternum segmentation.
• Developed a segmentation method using the closed Minimum Path over elliptical transformation of gradient images.
• Evaluated both methods on a dataset of 14 annotated MRI acquisitions.

Main Results:

• The gradient-based method demonstrated superior performance compared to K-Means clustering.
• The Gradient approach provided more accurate sternum segmentation in the evaluated MRI dataset.
• The study highlights the effectiveness of the Gradient method for sternum identification.

Conclusions:

• Automated sternum segmentation in MRI is feasible and beneficial for medical applications.
• The Gradient-based approach is a promising technique for accurate sternum segmentation in MRI.
• Further research can explore refining the Gradient method for enhanced clinical utility.