Post by @pangolin@sandbox.openscience.network

I am very happy with our publication "A human epithelial cell cycle dataset" which was published in Scientific Data.

In this study, we introduce a large-scale, high-resolution dataset consisting of 100,000 single-cell images of human HEp-2 cells. By combining expert manual annotations with deep learning (CNN) techniques, we have captured and classified the different phases of the cell cycle (from interphase to telophase).

The identification of mitotic figures is one of the most critical tasks in autoimmune diagnostics, particularly regarding the assessment of Antinuclear Antibodies (ANA). Because manual interpretation is often subject to human bias and morphological variability, our dataset provides a good resource for developing robust and smart algorithms. We think is a useful contribution toward standardizing and improving the precision of computer-aided diagnosis (CAD) of autoimmune diseases.

While we celebrate this scientific milestone, it is also a moment of deep gratitude and remembrance. We recently lost Karsten Conrad, who passed away at the age of 70. Privatdozent Dr. med. Karsten Conrad was until his retirement head of the Laboratory for Autoimmune Diagnostics at the Institute of Immunology, Carl Gustav Carus Faculty of
Medicine, Technical University Dresden.

Karsten was a mentor and a constant companion on my scientific journey. I remember clearly the early stages of my career in 2007/2008, when he supported me by publishing some my (first?) scientific documents in his book series. His guidance during those formative years was important to me.

I perceived Karsten as an incredibly active and prolific author, contributing numerous significant publications that have shaped the field. Beyond his research, he was a highly respected by everybody I know. I was happy to have at my conference, the International Biotech Innovation Days (IBID), later in career.

He was the driving force behind Association for the Advancement of Immune Diagnostics (GFID) (est. 2002) the world-renowned Dresden Symposium on Autoantibodies, which has served as a vital international platform for scientific exchange for 35 years.

I can certainly say that Karsten’s legacy is immense. His extensive body of work, his prolific writing and his dedication to research will surely continue to guide and benefit generations of scientists, physicians, and, most importantly, the patients who rely on this expertise.

Dear Karsten, thank you for your passion, your profound expertise, and your unwavering support.

#Immunology #AutoimmuneDiseases #ArtificialIntelligence #DeepLearning #Diagnostics #Biotechnology #ScientificData #InMemoriam

This infographic illustrates the methodology used to create the expert-annotated HEp-2 cell cycle dataset.
Visual Data: The top and bottom sections show grayscale microscopic images of HEp-2 cells categorized into seven distinct classes: Interphase, Prophase, Metaphase, Anaphase, Telophase, Artefact, and Triple. On the far right, simplified schematic diagrams illustrate the morphological characteristics of each phase.
The Workflow: The central flowchart details the creation process:
Image Acquisition:
1. Starting with HEp-2 microscopy images.
2. Cell Extraction: Individual cells are extracted from the images.
3. Blinded Annotation: Two independent experts perform blinded annotations on the extracted cells.
4. Conflict Resolution: The annotations are checked for disagreements, identical labels, or conflicting data.
5. Consensus: A joint assessment is performed to reach "agreed annotations.".
6. Final Output: This process results in the "Initial 7 classes dataset" containing 10,000 images. — This infographic illustrates the methodology used to create the expert-annotated HEp-2 cell cycle dataset. Visual Data: The top and bottom sections show grayscale microscopic images of HEp-2 cells categorized into seven distinct classes: Interphase, Prophase, Metaphase, Anaphase, Telophase, Artefact, and Triple. On the far right, simplified schematic diagrams illustrate the morphological characteristics of each phase. The Workflow: The central flowchart details the creation process: Image Acquisition: 1. Starting with HEp-2 microscopy images. 2. Cell Extraction: Individual cells are extracted from the images. 3. Blinded Annotation: Two independent experts perform blinded annotations on the extracted cells. 4. Conflict Resolution: The annotations are checked for disagreements, identical labels, or conflicting data. 5. Consensus: A joint assessment is performed to reach "agreed annotations.". 6. Final Output: This process results in the "Initial 7 classes dataset" containing 10,000 images.

OriginalPaper Scientific Data

A human epithelial cell cycle dataset

Rico Hiemann, Dirk Reinhold, Karsten Conrad, Stefan Rödiger, Peter Schierack, Dirk Roggenbuck September 1, 2026

doi:10.1038/s41597-026-08216-w

#Science, multidisciplinary #Science, Humanities and Social Sciences, multidisciplinary

We introduce a large-scale single cell dataset for cell cycle phases. The dataset includes single cell images of 4′,6-diamidino-2-phenylindole-stained human epithelioma-2 (HEp-2) cells grown on microscopic glass slides. Greyscale images of adherent HEp-2 cells were acquired at 20x magnification by an automated, inverse microscope system. Cell images were automatically segmented with 256 × 256 pixel patch size including surrounding area. Initially, 10,000 images were manually classified by experts into 5 cell-cycle groups (interphase, prophase, metaphase, anaphase, telophase) and two additional groups (artefact, triple). Based on these initial single cell-image classifications, a convolutional neural network (CNN) model (VGG16) was trained. Subsequently, additional segmented single cell images were pre-classified by the CNN and double-blinded reviewed as well as corrected by two experts to build an algorithm-aided dataset with 100,000 images. The dataset provides a resource for development of biomedical and deep learning applications and, thus, enables training and assessment of cell-cycle detection algorithms.

Read paper

No replies yet

Be the first to share your thoughts.

Open Science