A regularization technique for closed contour segmentation in ultrasound images

Area of Science:

• Medical Imaging
• Image Processing
• Computational Biology

Background:

• Ultrasound image segmentation is crucial for quantitative analysis of target object size and shape.
• Challenges include speckle noise, low signal-to-noise ratio (SNR), and low contrast, leading to boundary uncertainty.
• Conventional methods often fail to achieve accurate closed-curve segmentation due to weak edge detection.

Purpose of the Study:

• To propose a novel regularization model for improved ultrasound image segmentation.
• To enhance the ability to delineate target object boundaries as closed curves.
• To address limitations of conventional techniques in automatic segmentation.

Main Methods:

• Development of a new regularization model that controls curve smoothness to encourage closed curves.
• Integration of the model with various fitting terms to optimize segmentation.
• Validation through numerical simulations and experimental evaluations.

Main Results:

• The proposed model effectively encourages the formation of closed curves.
• Enhanced segmentation accuracy in extracting target object boundaries.
• Demonstrated improvement over conventional regularization techniques in overcoming noise and low contrast.

Conclusions:

• The novel regularization model significantly improves ultrasound image segmentation for closed-curve object delineation.
• The method offers a robust solution for challenging ultrasound imaging conditions.
• This approach advances quantitative analysis in medical imaging applications.