Adaptive color basis transformation. An aid in image segmentation

Area of Science:

• Biomedical image analysis
• Computational pathology
• Digital microscopy

Background:

• Multispectral imaging of stained cells allows segmentation and discrimination based on color differences.
• Constant spectral characteristics enable fixed linear combinations for image analysis.
• Non-constant spectral characteristics necessitate image-specific, often manual, analysis, hindering automation.

Purpose of the Study:

• To develop an automated algorithm for segmenting and discriminating stained cells in multispectral images.
• To address the challenge of non-constant spectral characteristics in cell images.
• To enable a fully automated process for cell analysis in digital pathology.

Main Methods:

• Utilizes principal components decomposition basis vectors for image analysis.
• Generates a non-static weighted linear combination of color images.
• Relies on a semiconstant relationship between the sizes of image components for segmentation.

Main Results:

• Successfully developed an automated algorithm for cell image segmentation.
• The algorithm effectively handles variations in spectral characteristics across images.
• Demonstrated successful application in segmenting stained cervical smears.

Conclusions:

• The developed algorithm provides an automated solution for cell segmentation and discrimination.
• Principal component analysis-based methods can overcome spectral variability in cell imaging.
• This technique aids in the automated analysis of stained cell images, such as cervical smears.