Deep Learning Techniques in Bone Marrow Cell Classification: A Review

Evaluating bone marrow cells is essential for diagnosing hematological disorders, the current manual procedure has many inherent problems. One significant hurdle is the complexity and variability of cell morphology, making accurate classification difficult even for experienced experts. Moreover, subjectivity introduces inconsistencies among individuals, leading to unreliable results. Manual classification is also time-consuming, constrained by limited training resources, and prone to intra-observer variability. The development of automated AI powered systems, standardized protocols and allocation of more resources are all key components of efforts to increase the accuracy and speed of bone marrow cell classification. The authors have evaluated and thoroughly analyzed previous studies on bone marrow cell classification that have used a variety of deep learning and machine learning approaches. This study provides an oversight of the existing deep learning and machine learning being used to aid and even improve traditional cytology.