Detection and Classification of Immature Leukocytes for Diagnosis of Acute Myeloid Leukemia Using Random Forest...

Area of Science:

• Hematology
• Computational Biology
• Medical Imaging

Background:

• Acute Myeloid Leukemia (AML) is a rapidly progressing, fatal blood cancer.
• Current AML diagnosis relies on manual peripheral blood smear examination, which is time-consuming, labor-intensive, and prone to inter-observer variability.
• There is a need for more efficient and objective diagnostic methods for AML.

Purpose of the Study:

• To develop and evaluate a machine learning model for the automated detection and classification of immature leukocytes for improved Acute Myeloid Leukemia (AML) diagnosis.
• To identify key image-based features that are discriminative for AML diagnosis.

Main Methods:

• Leukocyte images from AML patients and healthy controls were sourced from The Cancer Imaging Archive.
• Image processing techniques including format conversion, multi-Otsu thresholding, and morphological operations were employed for nucleus and cytoplasm segmentation.
• A Random Forest algorithm was trained using 16 extracted features, including two novel nucleus color features, for classification.

Main Results:

• The machine learning model achieved 92.99% accuracy for immature leukocyte detection and 93.45% accuracy for classification into four types.
• Precision values for each class exceeded 65%, representing an advancement over existing methods.
• The nucleus to cytoplasm area ratio was identified as a key discriminative feature, alongside two novel nucleus color features significant for classification.

Conclusions:

• The proposed machine learning model serves as a valuable support tool for the efficient diagnosis of Acute Myeloid Leukemia (AML).
• The identified discriminative features provide a foundation for future research in AML diagnostics.
• Automated analysis of leukocyte morphology shows promise for improving the accuracy and objectivity of AML diagnosis.