Leukocyte segmentation in peripheral blood images using a novel edge strength cue-based location detection method

Area of Science:

• Medical Imaging
• Computational Pathology
• Biomedical Engineering

Background:

• Accurate leukocyte classification aids in diagnosing blood disorders like leukemia and anemia.
• Effective segmentation of leukocytes from background is crucial for automated classification systems.
• Challenges in leukocyte segmentation include complex morphology, low contrast, and size/shape variations.

Purpose of the Study:

• To propose a novel framework for accurate leukocyte segmentation in peripheral blood images.
• To improve upon existing methods by introducing a new location detection mechanism prior to segmentation.
• To evaluate the performance and efficiency of the proposed leukocyte segmentation method.

Main Methods:

• A new framework combining leukocyte location detection and segmentation.
• Utilizing a novel edge strength cue (ESc) for initial leukocyte localization.
• Employing the Grabcut model for precise segmentation of identified leukocytes.
• Validation on diverse datasets: ALL-IDB1, Cellavision, and LISC.

Main Results:

• The proposed method demonstrates improved leukocyte segmentation accuracy.
• Experimental results evaluated using precision, recall, and F-score measures show superior performance.
• The method outperforms existing state-of-the-art techniques in leukocyte segmentation.
• Analysis of computation time indicates the method's efficiency.

Conclusions:

• The novel ESc-based location detection significantly enhances leukocyte segmentation accuracy.
• The proposed framework offers a robust and efficient solution for automated leukocyte analysis.
• This method holds potential for improving diagnostic accuracy in hematological diseases.