Artificial inteligence and datasets for leukemia diagnosis: A scoping review of machine lerning and deep learning approaches
Ashwini Tande, Renuka Mane · MethodsX · 2025
• Detailed domain knowledge required to understand blood formation process, leukemia, WHO classification standards and standard for diagnosis. • Detailed Analysis of publicly available benchmark datasets for leukemia diagnosis. • Review of ML and DL studies for leukemia detection and classification • Research gaps including dataset imbalance, WHO 2022 classification alignment Leukemia is the cancerous disease of the blood and the bone marrow that causes excessive proliferation of abnormal white blood cells, if detected too late, it can lead to potentially fatal consequences. Peripheral blood smear examinations and bone marrow biopsy are example of conventional diagnostic techniques that are invasive, time-consuming and subject to human variability. Recent advances in artificial intelligence (AI) particularly in the areas of Machine Learning (ML) and Deep Learning (DL)offer encouraging answers by making it possible to detect and classify leukemia using automated, effective and precise techniques. With an emphasis on image-based techniques based on publicly available datasets such as ALL-IDB, C-NMC, AML_Cytomorphology_LMU, SN-AM and CPTAC-AML, this paper reviews the most recent research on Machine Learning and Deep Learning approaches includes Convolutional Neural Networks (CNNS), ResNet, DenseNet, MobileNet and ensemble models for leukemia diagnosis. The survey highlights some of the most significant issues, such as dataset imbalance, stain variability, lack of standard annotations and limited clinical validation. The paper also discusses research gap and future initiatives including Explainable AI, lightweight deployment models, clinically reliable diagnostic system and hierarchical classification framework aligned with WHO 2022 classification standards.