This dataset comprises four prevalent AML subtypes with defining genetic abnormalities and typical morphological features according to the WHO 2022 classification: (i) APL with PML::RARA fusion, (ii) AML with NPM1 mutation, (iii) AML with CBFB::MYH11 fusion (without NPM1 mutation), and (iv) AML with RUNX1::RUNX1T1 fusion, as well as a control group of healthy stem cell donors. 

A total of

189 peripheral blood smears from the Munich Leukemia Laboratory (MLL) database from the years 2009 to 2020 were digitized

. First, all blood smears were scanned with 10x magnification and an overview image was created. Using the Metasystems Metafer platform, cell detection was performed automatically using a segmentation threshold and logarithmic color transformation. Further analysis regarding the quality of the region within the blood smear was performed automatically.

Per patient, 99-500 white blood cells

were then scanned in 40x magnification via oil immersion microscopy in .TIF format, corresponding to 24,9μm x 24,9μm (144x144 pixels). For this, a CMOS Color Camera from MetaSystems with a resolution of 4096x3000px and a pixel size of 3,45μm x 3,45μm was used. Four pixels were binned into one, leading to a size of 6.9μm x 6.9μm, and a resolution of 6.9μm / 40 (1px = 0,1725μm).

 Additional information about patient age, sex and blood counts are provided in a separate .csv file.

To our knowledge, this dataset covers the morphological complexity of acute myeloid leukemia in peripheral blood smears in unseen quality and quantity.