Self-initialized active contours for microscopic cell image segmentation.

Area of Science:

• Medical Image Analysis
• Computer Vision
• Computational Imaging

Background:

• Level set models excel at segmenting images with topological changes.
• Active contour models necessitate tedious manual parameter initialization.
• Image artifacts like intensity corruption challenge existing segmentation methods.

Purpose of the Study:

• To propose an incremental level set model with automatic contour initialization.
• To enhance image segmentation accuracy, especially in the presence of artifacts.
• To overcome limitations of manual parameter setting in active contour models.

Main Methods:

• Developed an incremental level set model utilizing local and global fitting energies for automatic contour initialization.
• Integrated region-based area and length terms with signed pressure force (SPF) to refine the segmentation process.
• Employed gradient descent flow for energy minimization, strengthened by SPF for smoother results.

Main Results:

• The proposed model demonstrates self-initialization, eliminating user intervention for parameter setting.
• Achieved higher accuracy in image segmentation compared to existing methods.
• Exhibited lower computational complexity and independence from initial contour placement.
• Validated superior performance on microscopic cell images against state-of-the-art models.

Conclusions:

• The novel incremental level set model offers an automated and accurate solution for image segmentation.
• Its robustness to artifacts and efficiency make it a valuable tool for medical image analysis.
• The model's self-initializing nature and performance advantages position it as an advancement in the field.