Deep Learning and Optimization Approaches in Segmentation and Classification of White Blood Cancer Cells in Bone Marrow Microscopic Images Using Deep Kronecker Neural Networks: A Review

Abstract

White blood cancer, particularly leukemia, is a life-threatening hematological disorder that requires early and accurate diagnosis for effective treatment. Bone marrow microscopic image analysis plays a critical role in detecting abnormal white blood cells; however, manual examination is time-consuming, subjective, and prone to human error. In recent years, deep learning techniques have significantly improved the automation of segmentation and classification tasks in medical imaging.

This review focuses on deep learning and optimization approaches for segmenting and classifying white blood cancer cells, with particular emphasis on Deep Kronecker Neural Networks (DKNNs). These networks leverage Kronecker product-based factorization to reduce model complexity while preserving feature representation, making them efficient for high-dimensional medical data. Advanced architectures such as U-Net, ResNet, DenseNet, and hybrid CNN-based models have demonstrated high accuracy in leukemia detection tasks. Additionally, optimization techniques including metaheuristic algorithms, transfer learning, and attention mechanisms further enhance model performance and generalization.

The review analyzes studies from 2020 to 2024, highlighting recent advancements, comparative performance, and challenges. Results indicate that hybrid and optimized deep learning models achieve classification accuracy above 98%. However, challenges such as data imbalance, computational complexity, and interpretability remain. Future research directions include lightweight architectures, explainable AI, and integration of multi-modal data for improved clinical decision support.