Human Zygotes at the 2PN Stage and the Same Embryos on Day 2 after ICSI: a Paired Image Dataset
Description
This dataset contains 776 brightfield images of 388 human embryos from 100 patients treated with intracytoplasmic sperm injection (ICSI) at the Infertility Therapy Center of Alzahra Hospital, Rasht, Iran, between September 2015 and October 2016. Each embryo was photographed twice: as a zygote at the two-pronuclear (2PN) stage, 16 to 17 h after injection, and again on day 2 of culture, 43 to 45 h after injection. The two images of each embryo are linked through laboratory records (labelled culture positions and the laboratory worksheet), not through visual similarity. Images were acquired with an Olympus IX70 inverted microscope and an Olympus DP71 CCD camera at ×150 and ×200 magnification. They are provided as BMP (264) or JPG (512) files in two folders, 2PN/ and Day2/. All images are 1280 × 1024 pixels except eight cropped images, whose dimensions are given in the metadata. Files are named P###_E##_2PN or P###_E##_D2, so the two images of an embryo share the same patient and embryo identifier. metadata.csv has one row per image, with the patient, embryo and image identifiers, the file name, the image dimensions and the embryologists' annotations: for 2PN images, the Z-score, off-centre position, off-centre juxtaposition, lack of juxtaposition, difference in pronuclear size and relative position of the pronuclei; for day-2 images, the number of blastomeres, the percentage of fragmentation and the embryo grade. patients.csv gives the serum β-hCG result of each patient's treatment cycle (45 positive, 55 negative). This result refers to the whole cycle, not to individual embryos, and should be analysed at the patient level. splits.csv assigns the patients to five predefined folds (20 patients and 77 or 78 embryos each), so that models can be evaluated without embryos of the same patient appearing in both training and test sets. make_splits.py regenerates the folds and can export the images and metadata of each fold into separate folders. README.md describes all files, variables and codes. The data can be used for developmental stage classification, detection and segmentation of pronuclei and blastomeres, automated pronuclear scoring and day-2 grading, and for studying how features at the 2PN stage relate to day-2 morphology. Patient identifiers are pseudonymised, and the images contain no identifying information. The study was approved by the ethics committee of Guilan University of Medical Sciences, and all patients gave written informed consent.
Files
Institutions
- University of GuilanGilan Province, Rasht