Skip to content

Latest commit

 

History

History
152 lines (121 loc) · 7.74 KB

BoneMarrowCytomorphology.md

File metadata and controls

152 lines (121 loc) · 7.74 KB

Bone Marrow Cytomorphology

Dataset Information

The Bone Marrow Cytomorphology MLL Helmholtz Fraunhofer dataset is a dataset focused on bone marrow cytomorphology. It contains over 170,000 anonymized, expert-annotated cell images from bone marrow smears of 945 patients, stained using May-Grünwald-Giemsa/Pappenheim. This dataset is not only the largest expert-annotated image pool of bone marrow cytomorphology currently available in the literature, but also its high quality and detailed classification of cell types make it an ideal choice for training leukocyte morphological classifiers. These classifiers can identify a range of diagnostically relevant cell types with high precision and recall.

Clinically, bone marrow cytomorphology examination is crucial for diagnosing various diseases, such as anemia, fever, and bleeding. By examining the morphology and quantity of cells in the bone marrow under a microscope, doctors can identify various blood diseases, such as leukemia, aplastic anemia, and multiple myeloma, and monitor changes in the disease during treatment, thus evaluating treatment effectiveness and prognosis. Classifying cells into 21 categories aids in more precise diagnosis and treatment because different types of cells play various roles in hematologic diseases. Their proportion and morphological changes can reflect the nature and progression of the disease. For example, an increased proportion of blast cells may suggest leukemia, while anomalies in megakaryocytes may be related to thrombocytopenic purpura or aplastic anemia.

Dataset Meta Information

Dimensions Modality Task Type Anatomical Structures Anatomical Area Number of Categories Data Volume File Format
2D Pathology Image Classification Blood Cells Marrow 21 171,375 JPG

Resolution Details

Dataset Statistics size
min [250,250]
median [250,250]
max [250,250]

Label Information Statistics

Abbreviation Category Chinese Equivalent Count
ABE Abnormal eosinophil 异常嗜酸细胞 8
ART Artefact 伪影 19630
BAS Basophil 嗜碱细胞 441
BLA Blast 原始细胞 11973
EBO Erythroblast 早红细胞 27395
EOS Eosinophil 嗜酸细胞 5883
FGC Faggott cell 费戈特细胞 47
HAC Hairy cell 毛细胞 409
KSC Smudge cell 污渍细胞 42
LYI Immature lymphocyte 未成熟淋巴细胞 65
LYT Lymphocyte 淋巴细胞 26242
MMZ Metamyelocyte 中幼粒细胞 3055
MON Monocyte 单核细胞 4040
MYB Myelocyte 骨髓细胞 6557
NGB Band neutrophil 带状中性粒细胞 9968
NGS Segmented neutrophil 分叶中性粒细胞 29424
NIF Not identifiable 无法识别 3538
OTH Other cell 其他细胞 294
PEB Proerythroblast 原始红细胞 2740
PLM Plasma cell 浆细胞 7629
PMO Promyelocyte 早幼粒细胞 11994

Proportion of each category.

Visualization

Example of each category.

File Structure

Images of different categories of bone marrow cells in the dataset are organized into subfolders, each containing a series of numbered image files.

Bone-Marrow-Cytomorphology
├── ABE
│ ├── ABE_00001.jpg
│ ├── ABE_00002.jpg
│ ├── ...
│ └── ABE_00008.jpg
├── ART
│   ├── 0001-1000
│   │   ├── ART_00001.jpg
│   │   ├── ART_00002.jpg
│   │   ├── ...
│   │   └── ART_01000.jpg
│   ├── 1001-2000
│   │   ├── ART_01001.jpg
│   │   ├── ART_01002.jpg
│   │   ├── ...
│   │   └── ART_02000.jpg
│   │   ├── ...
│   └── 19001-19630
│       ├── ART_19001.jpg
│       ├── ART_19002.jpg
│       ├── ...
│       └── ART_19630.jpg
├── ...
└── PMO
    ├── 0001-1000
    │   ├── PMO_00001.jpg
    │   ├── PMO_00002.jpg
    │   ├── ...
    │   └── PMO_01000.jpg
    ├── 1001-2000
    │   ├── PMO_01001.jpg
    │   ├── PMO_01002.jpg
    │   ├── ...
    │   └── PMO_02000.jpg
    │   ├── ...
    └── 11001-11994
        ├── PMO_11001.jpg
        ├── PMO_11002.jpg
        ├── ...
        └── PMO_11994.jpg

Authors and Institutions

Christian Matek（Institute of Computational Biology, Helmholtz Zentrum München, German; Department of Internal Medicine III, University Hospital Munich, Ludwig-Maximilians-Universität München, German; Institute of AI for Health, Helmholtz Zentrum München, German）

Sebastian Krappe（Image Processing and Medical Engineering Department, Fraunhofer Institute for Integrated Circuits IIS, German; Department of Computer Science, University of Koblenz-Landau, German）

Christian Munzenmayer（Image Processing and Medical Engineering Department, Fraunhofer Institute for Integrated Circuits IIS, German）

Torsten Haferlach（MLL Munich Leukemia Laboratory, German）

Carsten Marr（Institute of Computational Biology, Helmholtz Zentrum München, German; Institute of AI for Health, Helmholtz Zentrum München, German）

Source Information

Official Website: https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=101941770

Download Link: https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=101941770

Article Address: https://ashpublications.org/blood/article/138/20/1917/477932/Highly-accurate-differentiation-of-bone-marrow

Publication Date: 2021-11-08

Citation

@article{matek2021highly,
  title={Highly accurate differentiation of bone marrow cell morphologies using deep neural networks on a large image data set},
  author={Matek, Christian and Krappe, Sebastian and M{\"u}nzenmayer, Christian and Haferlach, Torsten and Marr, Carsten},
  journal={Blood, The Journal of the American Society of Hematology},
  volume={138},
  number={20},
  pages={1917--1927},
  year={2021},
  publisher={American Society of Hematology Washington, DC}
}

Original introduction article is here.