A Level Set Approach to Image Segmentation With Intensity Inhomogeneity.

Area of Science:

• Medical Imaging
• Image Processing
• Computer Vision

Background:

• Image segmentation is challenging with intensity inhomogeneity.
• Existing region-based methods fail due to reliance on intensity homogeneity.
• Bias fields in magnetic resonance images further complicate segmentation.

Purpose of the Study:

• To present a novel level set method for image segmentation.
• To address the challenge of intensity inhomogeneity in images.
• To enable simultaneous segmentation and bias correction.

Main Methods:

• Modeling inhomogeneous objects as Gaussian distributions with varying means and variances.
• Utilizing a sliding window to transform the image domain.
• Estimating bias fields adaptively by multiplying a bias field with the original signal.
• Defining a maximum likelihood energy functional incorporating bias field, level set function, and piecewise constant approximation.

Main Results:

• The proposed method effectively segments images with intensity inhomogeneity.
• Simultaneous segmentation and bias correction achieved for 3T and 7T MRI.
• Demonstrated superiority over existing representative algorithms on synthetic and real images.

Conclusions:

• The novel level set method offers improved image segmentation performance.
• The technique is robust in the presence of intensity inhomogeneity and bias fields.
• Applicable for advanced medical imaging tasks like MRI segmentation and bias correction.