Dataset Card for VISEM-Tracking

Dataset Summary

VISEM-Tracking is a dataset of human spermatozoa tracking, featuring 20 video recordings of 30 seconds each, captured from wet semen preparations. The dataset contains 29,196 frames with manually annotated bounding-box coordinates and a set of sperm characteristics analyzed by domain experts. In addition to the labeled dataset, unlabeled videos are provided to enable research in self-supervised learning. This dataset serves as an extension of the VISEM dataset, incorporating tracking data and additional clinical metadata.

Dataset Details

Curated by: Vajira Thambawita, Steven A. Hicks, Andrea M. Storås, Thu Nguyen, Jorunn M. Andersen, Oliwia Witczak, Trine B. Haugen, Hugo L. Hammer, Pål Halvorsen, Michael A. Riegler.
Funded by: Research Council of Norway under contract 270053.
Shared by: Simula Metropolitan Center for Digital Engineering, Oslo Metropolitan University.
Language(s) (NLP): Not applicable (biomedical imaging dataset).
License: Creative Commons Attribution 4.0 International (CC BY 4.0).

Dataset Sources

Repository: Zenodo - VISEM-Tracking
Paper: Nature Scientific Data - VISEM-Tracking
Code Repository: VISEM-Tracking GitHub

Uses

Direct Use

Sperm tracking and motility analysis for clinical and biomedical research.
Machine learning model training for spermatozoa detection and tracking.
Computer-aided sperm analysis (CASA) system development.
Deep learning-based sperm motility assessment.
Semi-supervised learning applications using unlabeled data.

Out-of-Scope Use

Not designed for direct clinical diagnosis.
Not suitable for general object detection tasks without domain-specific adaptation.

Dataset Structure

The dataset consists of:

20 annotated 30-second videos (45–50 FPS).
336 unlabeled 30-second videos for self-supervised learning.
Bounding box annotations (YOLO format) with tracking IDs.
Three sperm categories:
- 0: Normal sperm
- 1: Sperm clusters
- 2: Small or pinhead sperm
Metadata CSV files containing:
- Semen analysis results (e.g., sperm concentration, motility).
- Fatty acid composition (serum & spermatozoa).
- Sex hormone levels.
- Participant-related data.

Dataset Creation

Curation Rationale

Manual sperm motility assessment is challenging and time-consuming, requiring expert training. VISEM-Tracking was created to provide a large-scale, annotated dataset for training computer-aided sperm analysis (CASA) models, enabling automated sperm tracking and classification.

Source Data

Data Collection and Processing

Collected between 2008 and 2013 as part of a study on male reproductive function and obesity.
Videos captured at 400× magnification using an Olympus CX31 microscope with a UEye UI-2210C camera.
Bounding box annotations created using LabelBox, validated by domain experts.
Annotations include tracking IDs for spermatozoa over time.

Who are the source data producers?

Collected from 20 male participants, aged 18 years or older.
Approved by the Regional Committee for Medical and Health Research Ethics, South East, Norway (REK 2008/3957).
Participants provided written informed consent and data was anonymized.

Annotations

Annotation process

Bounding boxes were manually annotated by data scientists.
Three biologists verified annotations for accuracy and consistency.
Labels include sperm class, movement tracking, and cluster categorization.

Who are the annotators?

Researchers from Simula Metropolitan Center for Digital Engineering and Oslo Metropolitan University.
Experts in male reproduction and computer vision.

Personal and Sensitive Information

All participant data is anonymized.
No personally identifiable information is included.
Contains biological sample analysis data (e.g., hormone levels, fatty acid composition).

Bias, Risks, and Limitations

The dataset only includes 20 subjects, limiting its generalizability.
The dataset may not fully capture sperm motility variations across populations.
Annotations are limited to bounding boxes, without detailed morphological analysis.

Recommendations

Users should consider dataset limitations when applying it to real-world clinical applications.
Further data collection and augmentation are recommended for improving model generalization.

Citation

If you use this dataset, please cite:

BibTeX:

@article{thambawita2023visem,
  author = {Thambawita, Vajira and Hicks, Steven A. and Storås, Andrea M. and Nguyen, Thu and Andersen, Jorunn M. and Witczak, Oliwia and Haugen, Trine B. and Hammer, Hugo L. and Halvorsen, Pål and Riegler, Michael A.},
  title = {VISEM-Tracking, a human spermatozoa tracking dataset},
  journal = {Scientific Data},
  volume = {10},
  year = {2023},
  doi = {10.1038/s41597-023-02173-4}
}

More Information

For additional details, refer to:

Dataset Card Authors

Vajira Thambawita
Steven A. Hicks
Andrea M. Storås
Hugo L. Hammer
Michael A. Riegler

Dataset Card Contact

For inquiries, contact:

Vajira Thambawita ([email protected])