Sperm chromatin dispersion microscopy image dataset for instance segmentation

SCD Microscopy Image Dataset for YOLOv8-seg Instance Segmentation ================================================================= This repository contains a curated microscopy image dataset derived from the Sperm Chromatin Dispersion (SCD) assay, together with polygonal instance segmentation annotations and a YOLOv8s-seg benchmark model for sperm DNA fragmentation image analysis. The dataset was developed to support research on automated sperm cell instance segmentation and classification into fragmented and non-fragmented categories using bright-field microscopy images. Repository structure -------------------- SCD_Microscopy_YOLOv8seg_Repository/ | |-- README.txt|-- LICENSE.txt |-- scd_yolov8_seg.yaml | |-- dataset/ | |-- images/ | |-- images_split/ | | |-- train/ | | |-- val/ | | |-- test/ | | | |-- labels_split/ | |-- train/ | |-- val/ | |-- test/ | |-- annotations/ | |-- coco/ | |-- model/ | |-- best_yolov8s_seg_scd.pt| |-- final_selected_model/ Dataset description ------------------- The dataset consists of bright-field microscopy images obtained from SCD assay preparations. The images were organized using coded slide identifiers and manually annotated for instance segmentation. Dataset summary: - Total images: 2,000 PNG images - Image resolution: 1920 x 1080 pixels - Coded slides: 40 slides, labeled L01 to L40 - Annotated images: 1,994 images - Empty images: 6 images without annotable spermatozoa - Annotated sperm cells: 3,743 instances - Annotation type: polygonal instance segmentation - Annotation format: YOLOv8-seg and COCO - Classes: Fragmented and Non-fragmented The six images intentionally retained without annotations are: - L22_1083.png - L22_1092.png - L22_1105.png - L23_1143.png - L23_1149.png - L24_1189.png These images were preserved because no clearly annotable spermatozoa were present in the field of view. Classes ------- The dataset includes two classes: 0: Fragmented 1: Non-fragmented Dataset split ------------- The dataset was partitioned by coded slide to reduce image-level leakage between training, validation, and test subsets. - Train: 1,610 images - Validation: 200 images - Test: 190 images Annotation format ----------------- The YOLOv8-seg annotations are stored as text files. Each line represents one polygonal object instance using normalized coordinates: class_id x1 y1 x2 y2 ... xn yn where class_id corresponds to either fragmented or non-fragmented spermatozoa. Benchmark model --------------- The repository includes one selected benchmark model: model/best_yolov8s_seg_scd.pt This model corresponds to a YOLOv8s-seg instance segmentation model trained and selected for technical validation of the dataset. YOLO dataset configuration -------------------------- The YOLOv8 dataset configuration file is provided as: scd_yolov8_seg.yaml Installation ------------ Install the required Python dependencies using: pip install -r requirements.txt Basic YOLOv8 prediction example ------------------------------- from ultralytics import YOLO model = YOLO("model/best_yolov8s_seg_scd.pt") results = model.predict( source="dataset/images/test", imgsz=1024, conf=0.25, iou=0.50, retina_masks=True ) Intended use ------------ This dataset is intended for academic and scientific research in: - SCD microscopy image analysis - Sperm DNA fragmentation image interpretation - Instance segmentation of spermatozoa - Benchmarking segmentation models in reproductive biology and andrology imaging Limitations ----------- The dataset was generated from SCD bright-field microscopy images acquired under a specific laboratory workflow and imaging setup. External validation is recommended before applying models trained on this dataset to images acquired using different microscopes, staining conditions, cameras, laboratories, or image acquisition protocols. Authors ------- - Moises A. Adel Dominguez - Andres Mora Topete - Luis E. Magdaleno Marquez - Emiliano J. Ornelas Nande - Itzel Aranguren- Arturo Valdivia-G- Oscar D. Sanchez Affiliations ------------ Centro de Uro Andrologia, Guadalajara, Jalisco, Mexico. Universidad de Guadalajara, Centro Universitario de Ciencias Exactas e Ingenierias (CUCEI), Guadalajara, Jalisco, Mexico. Citation -------- If you use this dataset, annotations, code, or baseline models in your research, please cite the Zenodo repository below. Note that the associated Data Descriptor manuscript, “Sperm chromatin dispersion microscopy image dataset for instance segmentation”, is currently under peer review at Scientific Data (Nature Portfolio). Once published, citation details for the peer-reviewed article will be updated here. Ornelas Nande, E. J., Adel Domínguez, M. A., Mora Topete, A., Magdaleno Márquez, L. E., Aranguren, I., Valdivia-G, A., & Sanchez, O. D. (2026). Sperm chromatin dispersion microscopy image dataset for instance segmentation [Data set]. Zenodo. https://doi.org/10.5281/zenodo.21628868