Automatic leukocyte nucleus segmentation by intuitionistic fuzzy divergence based thresholding.

Area of Science:

• Medical Imaging
• Computational Biology
• Digital Pathology

Background:

• Accurate segmentation of leukocyte nuclei is crucial for blood cell analysis and disease diagnosis.
• Existing segmentation methods often struggle with noise and variations in microscopic blood smear images.

Purpose of the Study:

• To develop a robust automatic segmentation algorithm for leukocyte nuclei.
• To address challenges posed by noisy environments in microscopic blood smear analysis.

Main Methods:

• A novel exponential intuitionistic fuzzy divergence-based thresholding technique was developed.
• The algorithm minimizes image divergence using a new fuzzy entropy-based formula.
• A neighborhood-based membership function was designed to enhance noise handling.

Main Results:

• The algorithm achieved 98.52% average segmentation accuracy in noise-free environments.
• Under noisy conditions (Speckle and Gaussian), accuracies were 93.90% and 94.93%, respectively.
• The method demonstrated superior performance compared to existing algorithms, validated by expert hematologists.

Conclusions:

• The proposed fuzzy divergence technique offers a robust and accurate solution for leukocyte nucleus segmentation.
• The algorithm's effectiveness in noisy environments enhances its practical applicability in clinical diagnostics.
• This method shows significant potential for improving automated blood smear analysis.