Maximizing Kolmogorov Complexity for accurate and robust bright field cell segmentation.

Area of Science:

• Cellular imaging
• Microscopy
• Bioimage analysis

Background:

• Bright field defocused imaging offers low phototoxicity but lacks contrast for cell segmentation.
• Challenges include noise, experimental variability, and biological system unknowns.
• Fluorescent methods provide better contrast but involve higher phototoxicity.

Purpose of the Study:

• To develop a robust method for segmenting live cells in bright field defocused images.
• To improve cell tracking and analysis in microscopy.
• To overcome limitations of existing bright field imaging techniques.

Main Methods:

• Introduced maximal-information method using non-parametric information theory.
• Employed combinatorial optimization to select informative defocused images.
• Utilized Kolmogorov complexity to maximize set complexity.
• Applied selected images to initialize and guide a level set segmentation algorithm.

Main Results:

• Maximal-information method significantly improved segmentation precision and recall.
• Demonstrated superior performance over fixed defocused image selection strategies.
• Validated on diverse datasets of embryonic kidney cells (HEK 293T).

Conclusions:

• Combinatorial optimization and Kolmogorov complexity effectively extract information from bright field defocused images.
• The adaptive maximal-information approach is application-independent.
• Potential for processing noisy, high-throughput biological data.