A large dataset of bone marrow cells in myelodysplastic syndrome for classification systems

Bone marrow smear cytomorphology serves as the cornerstone for diagnosing hematological malignancies. It is particularly complex in myelodysplastic syndrome (MDS), which brings challenges to automated image analysis approaches. This dataset provides 25,009 images of the bone marrow cells from 125 patients with MDS. Image tags consist of 27 categories, including normal cells and abnormal cells in MDS that are characterized by pathological hematopoiesis, such as micro megakaryocytes. Each label underwent rigorous review by up to three independent experts. Cells were cropped after converting their bounding boxes into squares based on the longest side and expansion by 10% to guarantee the integrity of the cell boundaries and morphological features in the cropped images while minimum surrounding cells were included. Through training using this database, an AI model could attain the ability to differentiate dysplastic cells.