An ensemble-acute lymphoblastic leukemia model for acute lymphoblastic leukemia image classification

Area of Science:

• Medical Image Analysis
• Computational Biology
• Oncology

Background:

• Timely diagnosis of acute lymphoblastic leukemia (ALL) is critical for effective treatment and improved patient survival rates.
• Current diagnostic methods can be time-consuming, highlighting the need for enhanced early detection capabilities.
• Automated image classification holds potential to streamline the diagnostic process for medical practitioners.

Purpose of the Study:

• To introduce an ensemble-ALL model for the image classification of acute lymphoblastic leukemia (ALL).
• To enhance early diagnostic capabilities and streamline diagnostic and treatment processes.
• To improve the accuracy and efficiency of ALL diagnosis through advanced machine learning techniques.

Main Methods:

• Utilized a publicly available dataset partitioned into training, validation, and test sets.
• Employed and evaluated diverse convolutional neural networks (CNNs): InceptionV3, EfficientNetB4, ResNet50, CONV_POOL-CNN, ALL-CNN, Network in Network, and AlexNet.
• Developed an ensemble-ALL model by integrating top-performing SE-module-enhanced CNNs and optimizing with Bayesian optimization.

Main Results:

• The proposed ensemble-ALL model achieved high performance metrics: 96.26% accuracy, 96.26% precision, 96.26% recall, 96.25% F1-score, and 91.36% kappa score.
• Performance surpassed existing state-of-the-art studies in ALL image classification.
• Demonstrated the efficacy of the ensemble approach combined with SE modules and Bayesian optimization.

Conclusions:

• The ensemble-ALL model offers a significant advancement in the automated diagnosis of ALL via image classification.
• This model can enhance the efficiency and accuracy of medical professionals in diagnosing and treating ALL.
• Represents a valuable contribution to medical image recognition for acute lymphoblastic leukemia detection.