Automatic analysis system for abnormal red blood cells in peripheral blood smears

Area of Science:

• Hematology
• Medical Imaging
• Machine Learning

Background:

• Peripheral blood smear tests identify abnormal red blood cells (RBCs), crucial for diagnosing various diseases.
• Manual RBC classification is subjective, time-consuming, and lacks objectivity in clinical practice.

Purpose of the Study:

• To develop an automated machine learning (ML) method for accurate and objective RBC classification from microscopic images.
• To optimize classification criteria for RBC size and hemoglobin abnormalities.
• To classify morphologically abnormal RBCs using geometric features and multiple ML classifiers.

Main Methods:

• Utilized seven geometric features (e.g., major/minor axis, circularity) and five ML classifiers (SVM, Decision Tree, KNN, Random Forest, Adaboost).
• Optimized criteria for RBC size and hemoglobin abnormalities.
• Implemented a graphical user interface (GUI) for simultaneous classification and results display.

Main Results:

• Achieved highly accurate classification results (99.9%) using Support Vector Machine (SVM) for categorization.
• Demonstrated the feasibility of automated classification of morphologically abnormal RBCs.
• Provided simultaneous classification with results presented via a GUI.

Conclusions:

• The proposed ML-based automated method offers a highly accurate and objective alternative to manual RBC classification.
• This approach can significantly improve efficiency and consistency in clinical hematology diagnostics.
• The system provides real-time classification results through an intuitive GUI.