cAItomorph: Peripheral Blood Smear Image Dataset for Hematological Malignancy Prediction

Overview

This dataset represents the test set of the cAItomorph data, as used in the paper "AI-Based Hematological Malignancy Prediction from Peripheral Blood Smears in a Large Diagnostic Laboratory Cohort".

It includes 201,560 single-cell images from 409 patients spanning eight distinct hematologic conditions (seven disease classes and one healthy cohort). This data represents an isolated 20% testing split of a larger curated diagnostic laboratory cohort sourced from the Munich Leukemia Laboratory (MLL) between 2021 and 2022.

Diagnostic Classes

The dataset features single white blood cell images hierarchically grouped from 168 highly specific labels down to 19 detailed classes, and finally into the 8 coarse classes provided in this release:

Acute leukemia 

Lymphoma 

Myelodysplastic syndromes (MDS)

MDS/MPN overlap syndromes

Myeloproliferative neoplasms (MPN)

Plasma cell neoplasms

Reactive changes

Healthy donors

Data Structure and Formatting

The dataset is organized to facilitate immediate use in machine learning pipelines, structured as follows:

Images: All images are cropped single white blood cells, stored in TIF format, with uniform dimensions of 144x144 pixels.

/dataset_root/
├── metadata.csv
└── patients/
├── ALK_184/
│ ├── Gal-000001.RGB.TIF
    │ └── Gal-000002.RGB.TIF
    ├── ALK_185/
└── ... (409 patient folders total)

Metadata Details (metadata.csv)

Column
Description
Examples

patient_id
Unique identifier matching the folder name.
ALK_184, ALK_185

diagnosis_coarse
Broad diagnostic categories.
Acute leukemia, MDS, Healthy

diagnosis_fine
Detailed diagnostic categories.
AML, B-cell neoplasm, CMML