Back

HSTLI, a Dataset of Human Semen Time-lapse Images for Detection, Tracking, and Motility Parameters Analysis

Sivri, A.; Choi, J.; Bopp, J.; Anouna, A.; VerMilyea, M.; Alkhoury, G.; Hocaoglu, O. O.; Kam, M.; Alkhoury, L.

2025-12-17 bioengineering
10.64898/2025.12.15.694470 bioRxiv
Show abstract

We present HSTLI, a dataset of Human Semen Time-Lapse Images acquired using two microscopy systems. These were (i) a commercial Computer-Assisted Semen Analysis (CASA) system and, (ii) an optical microscope. The dataset contains samples from 51 healthy male participants and includes 3,266 video clips (approximately 27 hours of imagery). A subset of the clips from both systems was manually annotated with bounding boxes around each visible sperm head, thereby establishing ground truth for detection, tracking, and motility analysis. Specifically, 14 CASA clips and 20 optical-microscope clips (each consisting of 900 frames) were labeled, yielding 29,950 annotated frames and roughly 1.4 million sperm annotations. The dataset also includes thousands of unlabeled clips captured under varying preparation conditions (washed vs. unwashed), dilution levels, magnifications, and regions of interest. When available, motility parameter reports are provided from either clinical technicians or CASA outputs. We demonstrate the use of this dataset in two example applications, (i) automated sperm detection and tracking using YOLOv5 and SORT, and (ii) visualizations of motility parameters. HSTLI offers a comprehensive benchmark for developing and evaluating algorithms for sperm detection, tracking, classification, and motility assessment.

Matching journals

The top 10 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.