Systematic segmentation method based on PCA of image hue features for white blood cell counting

Area of Science:

• Medical image analysis
• Computational pathology
• Hematology

Background:

• Accurate leukocyte (white blood cell, WBC) counts are crucial for diagnosing infections and guiding treatment.
• Current manual or semi-automated WBC counting methods are prone to miscounting.
• Automated methods are needed to improve the accuracy and efficiency of WBC quantification.

Purpose of the Study:

• To develop an automated, bioinspired segmentation method for precise leukocyte counting in blood smear images.
• To leverage chromatic contrast, mimicking human visual perception, for enhanced WBC identification.
• To improve the accuracy and efficiency of WBC quantification compared to existing methods.

Main Methods:

• A bioinspired segmentation approach using principal component analysis across RGB, HSV, and L*a*b* color spaces.
• Utilizing high chromatic variance of leukocyte pixels for increased contrast against other blood elements.
• Processing chromaticity as a feature without separating hue components, enhancing computational efficiency.

Main Results:

• The proposed method demonstrated computationally efficient WBC recognition and separation of overlapping cells.
• Achieved high specificity (approx. 0.2% higher) and accuracy (approx. 0.3% higher) in pixel classification compared to previous methods.
• WBC counting results were comparable to existing techniques, with improved segmentation precision.

Conclusions:

• The developed automated method offers a precise and efficient approach to leukocyte counting in blood smears.
• The bioinspired segmentation technique effectively utilizes chromatic contrast for accurate WBC identification.
• This method has the potential to improve diagnostic accuracy and treatment monitoring in clinical settings.