Correction tool for Active Shape Model based lumbar muscle segmentation.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Accurate and rapid image segmentation is crucial for analyzing pathological regions in clinical settings.
• Existing anatomical image segmentation methods often require time-consuming manual corrections.
• There is a need for efficient tools that minimize user interaction for segmentation refinement.

Purpose of the Study:

• To develop an interactive correction method for improving 3D image segmentation results.
• To enable faster and more intuitive segmentation corrections with minimal user input.
• To evaluate the effectiveness of the proposed method for lumbar muscle segmentation.

Main Methods:

• A novel interactive correction method utilizing a 2D/3D environment for 3D shape correction.
• Direct manipulation of free-form deformation adapted for 2D interactions.
• Implementation into a software tool and evaluation on Magnetic Resonance Images (MRIs) for lumbar muscle segmentation.

Main Results:

• The developed tool allows for intuitive 3D segmentation correction via simple 2D interactions.
• Full segmentation correction was achieved in an average of 6±4 minutes with 68±37 interactions.
• High segmentation quality was maintained, evidenced by an average Dice coefficient of 0.92±0.03.

Conclusions:

• The proposed interactive correction method offers an efficient and user-friendly solution for refining 3D image segmentations.
• The tool significantly reduces correction time and user effort while preserving segmentation accuracy.
• This approach holds promise for improving clinical workflows in anatomical image analysis.