Automated Bone Marrow Cell Classification for Haematological Disease Diagnosis Using Siamese Neural Network

Area of Science:

• Hematology
• Medical Imaging
• Machine Learning

Background:

• Accurate classification of bone marrow cells is crucial for diagnosing hematological disorders.
• Manual classification is time-consuming and prone to human error, even for experts.
• Automating this process can lead to faster and more accurate diagnoses.

Purpose of the Study:

• To develop and evaluate machine learning models for automated classification of bone marrow cells.
• To improve the speed and accuracy of diagnosing hematological ailments through cell structure analysis.
• To compare the performance of different machine learning algorithms for this task.

Main Methods:

• A dataset of 170,000 expert-annotated bone marrow cell images from 945 patients was utilized.
• Several machine learning models were trained and tested, including Convolutional Neural Network (CNN) + Support Vector Machine (SVM), CNN + XGBoost, and a Siamese network.
• Model performance was evaluated using accuracy, precision, and recall metrics.

Main Results:

• CNN + SVM and CNN + XGBoost models achieved low accuracies of 32% and 28%, respectively.
• The Siamese neural network demonstrated superior performance with 91% accuracy and 84% validation accuracy.
• The Siamese network achieved weighted average recall values of 92% for training and 91% for validation.

Conclusions:

• The Siamese neural network model significantly outperforms other tested algorithms for bone marrow cell classification.
• Automated classification using the Siamese network offers a faster and more accurate approach to diagnosing hematological disorders.
• This AI-driven method has the potential to enhance diagnostic efficiency in hematology.