Quantum-Inspired Deep Feature Selection and Transfer Learning Approach for Leukemia Disease

Abstract

Leukemia is a critical blood malignancy requiring early and precise diagnosis for effective treatment. Leveraging advancements in artificial intelligence (AI) and quantum computing, this study presents an innovative methodology for leukemia detection using deep feature extraction and quantum-enhanced feature selection. Convolutional Neural Networks (CNNs) were utilized for extracting high-dimensional features from blood smear images, capturing complex morphological patterns. To address the challenges of high dimensionality and redundant data, a Quantum-Inspired Optimization Algorithm (QIOA) was implemented for feature selection, significantly optimizing the input feature space for machine learning models. Using the optimized features, the study assessed six machine learning classifiers: Naive Bayes, K-Nearest Neighbors (KNN), Random Forest, Decision Tree, Support Vector Machines (SVM), and Logistic Regression. While Decision Tree and SVM showed good generalization skills, Logistic Regression had the highest accuracy of 96.72% with balanced precision and recall. This method demonstrates how AI and quantum computing concepts might be combined to enhance medical diagnosis. The findings open the door for future applications in intricate medical datasets by proving the viability and efficacy of quantum-enhanced machine learning for precise and effective leukemia identification.