Classification of chronic myeloid leukemia cell subtypes based on microscopic image analysis

Area of Science:

• Medical Imaging
• Hematology
• Computational Biology

Background:

• Chronic Myeloid Leukemia (CML) diagnosis relies on accurate cell classification.
• Microscopic analysis of bone marrow and peripheral blood smears is crucial for CML detection.
• Automated methods can enhance the efficiency and reliability of CML diagnosis.

Purpose of the Study:

• To develop a simple, efficient computer-aided diagnosis (CAD) method for classifying Chronic Myeloid Leukemia (CML) cells.
• To introduce a novel feature set and decision tree classifier for CML cell categorization.
• To evaluate the performance of the proposed CAD method against expert manual labeling.

Main Methods:

• Utilized microscopic image processing for CML cell classification.
• Introduced a novel combination of typical and new features for classification.
• Employed an effective decision tree classifier to categorize CML cells into eight groups.
• Evaluated the method on 1730 CML cell images (714 non-cancerous, 1016 cancerous).

Main Results:

• Achieved average accuracy of 99.0%, specificity of 99.4%, and sensitivity of 98.3% for CML cell classification.
• Demonstrated high conformity (Cohen's kappa coefficient of 0.99) with expert diagnoses.
• The method showed high capability in classifying CML cells effectively.

Conclusions:

• The proposed computer-aided diagnosis method is a simple, affordable, and reliable tool for CML diagnosis.
• The method assists pathologists by providing accurate classification of CML cells.
• Image processing and decision tree classification offer a promising approach for hematological malignancy diagnosis.