OocyHistDB: A dataset of histological images for oocyte detection and segmentation.

# OocyHistDB: Histological Image Dataset for Oocyte Analysis in *Centropomus undecimalis* ## Overview **OocyHistDB** is a histological image dataset developed to support computational studies on the detection and segmentation of oocyte developmental stages in the fish species *Centropomus undecimalis* (common snook). The dataset was created to provide a structured and publicly available resource for researchers working with computer vision, deep learning, fish reproductive biology, histology, and digital image analysis. The dataset contains high-resolution RGB histological images of ovarian tissue, with expert annotations representing different oocyte developmental stages. It is intended to serve as a benchmark for the development, evaluation, and comparison of machine learning and deep learning methods for histological image analysis. This dataset was originally introduced in the article **“OocyHistDB: A histologic image dataset for detection oocytes”**, published in *Revista de Sistemas e Computação*, v. 12, n. 3, p. 61–68, 2022, DOI: **10.36558/rsc.v12i3.7945**. ## Biological Context Oocyte development analysis is an important step in reproductive studies of fish species. In *Centropomus undecimalis*, the characterization of ovarian structures can contribute to studies on reproductive biology, fishery management, and conservation of aquatic resources. Histological analysis traditionally depends on manual inspection by specialists, which can be time-consuming and subject to observer variability. Therefore, annotated image datasets such as OocyHistDB are valuable for developing automated computational tools capable of assisting specialists in the identification and quantification of oocyte developmental stages. ## Dataset Description The dataset contains histological images of ovarian tissue from *Centropomus undecimalis*. The images were acquired from histological sections stained with Hematoxylin and Eosin (H&E), using optical microscopy. The dataset includes annotations for three oocyte developmental classes: | Class ID | Class abbreviation | Description ||---|---|---|| 0 | PV | Pre-vitellogenesis || 1 | VI | Initial vitellogenesis || 2 | VF | Final vitellogenesis | The pre-vitellogenesis class groups early oocyte stages, while VI and VF represent more advanced vitellogenic stages. ## Image Acquisition The histological images were obtained from ovarian tissue sections of *Centropomus undecimalis*. The tissue samples were processed using standard histological procedures and stained with Hematoxylin and Eosin. Images were captured using a LEICA DM 500 microscope at 200× magnification, coupled to a LEICA EC4 digital camera. Each original image is an RGB image with a resolution of **640 × 640 pixels**. ## Dataset Size and Annotation The original image collection contained **305 histological images**. After data augmentation, the dataset was expanded to **772 images**. A total of **5,680 oocytes** were annotated by a specialist, distributed as follows: | Class | Number of annotated oocytes ||---|---:|| PV | 3,066 || VI | 782 || VF | 1,832 || **Total** | **5,680** | ## Dataset Split The dataset was divided into training, validation, and test subsets: | Subset | Number of images | PV | VI | VF | Total annotated oocytes ||---|---:|---:|---:|---:|---:|| Train | 708 | 2,393 | 633 | 1,500 | 4,526 || Validation | 32 | 299 | 51 | 139 | 489 || Test | 34 | 374 | 98 | 193 | 665 || **Total** | **772** | **3,066** | **782** | **1,832** | **5,680** | ## Annotation Format The dataset may be provided in formats commonly used for computer vision tasks, such as: - COCO JSON- YOLO format- Pascal VOC XML- CSV or TXT annotation files Depending on the released version, annotations may represent bounding boxes or segmentation masks/polygons. Users should check the specific annotation files included in this Zenodo record before training or evaluating models. ## Data Augmentation The augmented version of the dataset was generated using image transformations to increase data diversity and improve model robustness. The augmentation procedures included: - Horizontal and vertical flipping- Rotation by 90 degrees- Random cropping- Random rotation These transformations were applied while preserving the correspondence between images and annotations. ## Suggested Applications OocyHistDB can be used for several research tasks, including: - Oocyte detection- Semantic segmentation of oocyte classes- Instance segmentation of oocytes- Classification of oocyte developmental stages- Evaluation of deep learning models in histological images- Benchmarking computer vision methods for fish reproductive biology ## Recommended Citation If you use this dataset, please cite the original publication: ```bibtex@article{cruz2022oocyhistdb, title={OocyHistDB: A histologic image dataset for detection oocytes}, author={Cruz, Yanna Leidy Ketley Fernandes and Santana, Ewaldo Eder Carvalho and Silva, Antonio Fhillipi Maciel and Nascimento, Isa Rosete Mendes Araujo and Carvalho Neta, Raimunda Nonata Fortes}, journal={Revista de Sistemas e Computação}, volume={12}, number={3}, pages={61--68}, year={2022}, doi={10.36558/rsc.v12i3.7945}} URL dowload dataset: https://app.roboflow.com/ds/yUMSxVF6c5?key=5o8R6r9gvB