Transfer Learning for Multiclass Classification of Bone Marrow Cells
Rishabh Hanselia, Dilip Kumar Choubey, Kanchan Bala, Ashutosh Mishra · 2024
Accurate classification of bone marrow cells for hematological disorder diagnosis faces challenges due to complex morphology and subjectivity in manual assessment. Inconsistencies among experts, time constraints, and intra-observer variability further hinder the process. To address these issues, researchers are working on standardized criteria, improved resources, and AI-based automation. This study rigorously analyzes diverse deep and machine learning techniques in bone marrow cell classification, highlighting their role in modernizing cytology. The authors present their own CNN-based model for digital cytology, achieving a notable 91.66% test accuracy. The study is structured into five main sections: introduction, motivation, related work, material and methods, experimental results, and discussions along with future directions. This research showcases the potential of AI-driven approaches to enhance accuracy and efficiency in bone marrow cell classification, revolutionizing hematological disorder diagnostics.