High-Power Field Peripheral Blood Smear Images for Leukemia Classification (AML, ALL, CLL, CML, Reactive, Normal)

This dataset contains 2,832 high-power field (HPF) microscopy images of Wright–Giemsa–stained peripheral blood smears, distributed across five diagnostic categories:Acute Myeloid Leukemia (AML), Acute Lymphoblastic Leukemia (ALL), Chronic Myeloid Leukemia (CML), Chronic Lymphocytic Leukemia (CLL), and Normal/Reactive. APL (APML) images are included within the AML folder.Reactive lymphocytosis and reactive morphological fields are included within the Normal folder. All images were captured at 100× oil immersion and represent morphologically characteristic fields typically encountered in routine hematology practice. Annotation: All images in this dataset were independently reviewed and annotated by three experienced hematopathologists, ensuring high diagnostic accuracy and consistency. These annotated images formed the basis for model development in the following publication: Syed Naveed et al. Novel hierarchical deep learning models predict type of leukemia from whole-slide microscopic images of peripheral blood. Journal of Medical Artificial Intelligence (JMAI), 2024.https://jmai.amegroups.org/article/view/9379/html Role of This Dataset in Model Development: This dataset served as the training and internal testing dataset for developing the hierarchical deep learning models described in the above publication.The images represent clean, representative diagnostic fields of AML, ALL, CML, CLL, and normal/reactive blood films. Validation Dataset (Not Included Here) For external validation in the study, a separate, newly collected dataset was used.The validation dataset consisted of: All HPF fields from entire slides, Mixed content fields including normal cells, leukemic blasts, reactive patterns, Images that occasionally resembled different leukemia types—reflecting true real-world diagnostic complexity Although the model was trained on the structured dataset provided here, it successfully generalized to all smear images, demonstrating robustness and practical applicability. The validation set is not included in this public release. Intended Use: This dataset is made available for: Machine learning model training Morphological research Educational and academic use Benchmarking automated leukemia classification systems License and Citation This dataset is shared for non-commercial research purposes only. Researchers using this dataset are requested to cite: Syed Naveed et al. (2024). Novel hierarchical deep learning models predict type of leukemia from whole-slide microscopic images of peripheral blood. Journal of Medical Artificial Intelligence.https://jmai.amegroups.org/article/view/9379/html