Deep Learning Neural Network Based on PSO for Leukemia Cell Disease Diagnosis from Microscope Images

Area of Science:

• Medical Imaging
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Leukemia is a cancer of abnormal white blood cells (WBCs) originating in bone marrow.
• Accurate and timely diagnosis is crucial for effective leukemia treatment.
• Identifying leukemia cells in microscopic images presents significant challenges due to complex features.

Purpose of the Study:

• To develop a deep learning model for diagnosing leukemia from microscopic images.
• To optimize feature selection using metaheuristic algorithms for improved diagnostic accuracy.
• To evaluate the performance of different machine learning classifiers in conjunction with the proposed method.

Main Methods:

• Utilized deep learning for initial feature extraction from leukemia images.
• Employed Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) for feature selection.
• Applied Decision Tree (DT), Support Vector Machine (SVM), and K-Nearest Neighbors (K-NN) for classification.
• Tested models with GoogLeNet and ResNet-50 architectures.

Main Results:

• PSO with GoogLeNet achieved accuracies of 97.4% (SVM), 92.3% (K-NN), and 85.9% (DT).
• ACO with ResNet-50 yielded superior accuracies of 100% (SVM), 94.9% (K-NN), and 92.3% (DT).
• The proposed deep learning and optimization approach demonstrated high diagnostic performance.

Conclusions:

• The developed deep learning approach with optimized feature selection shows significant promise for accurate leukemia diagnosis.
• This method offers a potential tool for automated analysis of microscopic images for cancer detection.
• Further validation could enhance its clinical applicability in early leukemia detection.