Bone Marrow Cells Detection: A Technique for the Microscopic Image Analysis.

Area of Science:

• Biomedical Engineering
• Computational Pathology
• Hematology

Background:

• Accurate bone marrow cell (BMC) counting is crucial for diagnosing myeloproliferative diseases.
• Existing methods for BMC analysis can be labor-intensive and prone to variability.

Purpose of the Study:

• To develop and validate a novel automated method for localizing, segmenting, and classifying bone marrow cells.
• To improve the accuracy and efficiency of BMC analysis for myeloproliferative disease detection.

Main Methods:

• A three-module approach combining localization, segmentation (nucleus and cytoplasm), and classification.
• Utilized color transformation, stepwise averaging, Otsu's method, region growing, K-Means, and marker-controlled watershed algorithms.
• Employed Support Vector Machine (SVM) for classifying five BMC types: neutrophilic split granulocyte, neutrophilic stab granulocyte, metarubricyte, mature lymphocytes, and outliers.

Main Results:

• Achieved an average segmentation accuracy of 91.76%.
• Obtained an average recall rate of 87.49% for cell classification.
• Demonstrated superior performance compared to existing methods in both segmentation and classification.

Conclusions:

• The proposed automated method offers a significant advancement in bone marrow cell analysis.
• This technique enhances diagnostic capabilities for myeloproliferative diseases through improved accuracy and efficiency.