Ultrasound image segmentation with shape priors: application to automatic cattle rib-eye area estimation

Area of Science:

• Medical Imaging
• Image Processing
• Computational Anatomy

Background:

• Ultrasound (US) image segmentation is challenging due to noise and data limitations.
• Accurate segmentation is crucial for quantitative analysis in various medical applications.
• Existing methods struggle with the inherent variability and artifacts in US images.

Purpose of the Study:

• To develop an automated method for ultrasound image segmentation.
• To improve the accuracy of rib-eye shape segmentation using prior knowledge.
• To create a robust technique applicable to similar ultrasound imaging challenges.

Main Methods:

• A novel approach combining shape priors and US image information.
• Utilizing expert-segmented images to establish a mean rib-eye shape model.
• Employing a closed curve fitting method incorporating image data and geodesic distance to the mean shape.

Main Results:

• Successfully automated the segmentation of rib-eye shapes in US images.
• Demonstrated robustness against noise and information gaps.
• Validated on a dataset of 610 US images with expert-level accuracy.

Conclusions:

• The proposed method effectively segments rib-eye shapes in ultrasound images.
• This approach offers a significant advancement for automated US image analysis.
• The technique is adaptable for similar segmentation tasks in other medical fields.