A segmentation method of lung cavities using region aided geometric snakes

Area of Science:

• Medical Imaging
• Radiotherapy Planning
• Computational Anatomy

Background:

• Accurate lung segmentation in medical images is crucial for applications like radiotherapy.
• Challenges include varying lung shapes, low contrast, and ill-defined boundaries, complicating automated segmentation.
• Existing methods struggle with complex geometries and weak edge detection.

Purpose of the Study:

• To develop and evaluate a modified geometric-based snake model for improved automatic lung cavity segmentation in magnetic resonance (MR) images.
• To enhance segmentation efficiency in capturing complex geometries and addressing initialization and edge detection difficulties.
• To provide a robust tool for oncological applications, particularly radiotherapy treatment planning.

Main Methods:

• Utilized a modified geometric-based snake model integrating gradient flow forces and fuzzy c-means clustering for region constraints.
• The model was designed to improve segmentation efficiency for complex geometries and weak edges.
• Tested on a database of 30 MR images (80 slices each).

Main Results:

• The modified snake model demonstrated improved segmentation efficiency and robustness.
• Achieved encouraging results when compared to manual segmentations by an expert radiologist.
• Outperformed previous segmentation methods in handling challenging lung cavity characteristics.

Conclusions:

• The proposed modified snake model offers a robust and efficient approach for automatic lung segmentation in MR images.
• This method shows significant potential for improving radiotherapy treatment planning in oncology.
• The integration of gradient flow and fuzzy c-means clustering effectively addresses segmentation challenges.