Acute Lymphoblastic Leukemia Detection and Classification of Its Subtypes Using Pretrained Deep Convolutional Neural...

Area of Science:

• Medical Imaging
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Leukemia is a critical white blood cell disease impacting blood and bone marrow.
• Accurate detection and subtype classification of acute lymphoblastic leukemia (ALL) are vital for effective treatment.
• Previous research has often overlooked the classification of ALL subtypes.

Purpose of the Study:

• To develop and evaluate a deep convolutional neural network (CNN) for automated ALL detection.
• To classify ALL subtypes (L1, L2, L3) and normal cells using a fine-tuned pretrained model.
• To assess the performance of the model across different color models without requiring image segmentation.

Main Methods:

• Utilized a pretrained AlexNet model, fine-tuned on a custom dataset for ALL detection and classification.
• Implemented data augmentation techniques to mitigate overtraining and enhance model robustness.
• Compared model performance using various color models to evaluate effectiveness on diverse image types.

Main Results:

• Achieved 100% sensitivity, 98.11% specificity, and 99.50% accuracy for ALL detection.
• Attained 96.74% sensitivity, 99.03% specificity, and 96.06% accuracy for ALL subtype classification.
• Demonstrated high diagnostic performance without the necessity of microscopic image segmentation.

Conclusions:

• The proposed deep learning approach offers a highly accurate and efficient method for automated ALL detection and subtype classification.
• Fine-tuning pretrained networks like AlexNet provides a powerful strategy for medical image analysis tasks.
• The model's ability to perform accurately without segmentation simplifies the diagnostic workflow and broadens its applicability.