A large expert-annotated single-cell peripheral blood dataset for hematological disease diagnostics

Sayedali Shetab Boushehri ^1,2,3, Salome Kazeminia ^1,3, Armin Gruber ^1,4

¹Computational Health Center, Helmholtz Munich - German Research Center for Environmental Health, Neuherberg, Germany.
²Data & Analytics, Pharmaceutical Research and Early Development (pRED), Roche Innovation Center Munich (RICM), Penzberg, Germany.
³TUM School of Computation, Information and Technology, Technical University of Munich, Munich, Germany.
⁴Department of Medicine III, University Hospital, LMU Munich, Munich, Germany.
⁵German Cancer Consortium (DKTK), Heidelberg, Germany.
⁶German Cancer Research Center (DKFZ), Heidelberg, Germany.
⁷Munich Leukemia Laboratory, Munich, Germany.
⁸Computational Health Center, Helmholtz Munich - German Research Center for Environmental Health, Neuherberg, Germany. carsten.marr@helmholtz-munich.de.
⁹Department of Medicine III, University Hospital, LMU Munich, Munich, Germany. carsten.marr@helmholtz-munich.de.
¹⁰German Cancer Consortium (DKTK), Heidelberg, Germany. carsten.marr@helmholtz-munich.de.
¹¹Munich Center for Machine Learning (MCML), Munich, Germany. carsten.marr@helmholtz-munich.de.

⁰Computational Health Center, Helmholtz Munich - German Research Center for Environmental Health, Neuherberg, Germany.

Scientific Data +

|

November 11, 2025

View abstract on PubMed

Summary

Researchers developed a large, annotated dataset of over 40,000 peripheral blood cell images. This resource aids artificial intelligence in classifying blood diseases like leukemia, improving diagnostic tools.

Area Of Science

Hematology
Medical Diagnostics
Artificial Intelligence in Medicine

Background

Accurate classification of peripheral blood cells is crucial for diagnosing hematological malignancies, including leukemia subtypes.
Automating cell classification using artificial intelligence (AI) holds significant promise for improving diagnostic efficiency.
The development of robust AI algorithms necessitates large, high-quality, and well-annotated single-cell datasets.

Purpose Of The Study

To introduce and provide public access to a comprehensive, annotated dataset of peripheral blood cell images.
To facilitate the training and validation of machine learning models for automated hematological cell classification.
To support the advancement of reliable diagnostic tools for blood diseases.

Main Methods

Compilation of a large dataset (>40,000 images) of peripheral blood single-cell images.
Expert classification of images into 18 distinct cell types by cytomorphology specialists at the Munich Leukemia Laboratory.
Public release of the annotated dataset for research purposes.

Main Results

Creation of a >40,000 image, publicly available, annotated peripheral blood cell dataset.
Dataset classified into 18 cell types by expert cytomorphologists.
Establishment of a valuable resource for medical and machine learning research.

Conclusions

The released dataset is a significant resource for developing and validating AI-based diagnostic tools for hematological diseases.
Public availability of this data will accelerate research in automated cell classification and improve clinical diagnostics.
This resource supports the development of more reliable and clinically relevant diagnostic solutions for blood disorders.