Abstract
SEED-ML (Semen Examination and Evaluation Dataset for Machine Learning) is an openly available, multi-parametric clinical dataset specifically designed to support research in male infertility diagnostics and prediction. SEED-ML refers specifically to the dataset repository and its clinical structure, and not to a specific machine learning model or diagnostic method. In this sense, SEED-ML comprises records from 10,124 patients, including detailed semen analysis parameters (pre- and post-capacitation), morphological classifications, and clinical alterations. Infertility diagnosis is categorized into nine clinically relevant classes, ranging from normal fertility to complex multi-factor conditions such as oligoasthenoteratozoospermia. All data were anonymized and curated following strict ethical and privacy guidelines to ensure compliance with applicable medical data protection regulations. The dataset reflects real-world clinical distributions across nine diagnostic classes: Normozoospermia (62.68%), Oligoasthenoteratozoospermia (14.22%), Asthenozoospermia (11.66%), Teratozoospermia (6.71%), Oligozoospermia (1.90%), Asthenoteratozoospermia (1.38%), Oligoasthenozoospermia (0.96%), Oligoteratozoospermia (0.34%), and Azoospermia (0.16%). This detailed categorization provides a realistic clinical distribution for machine learning evaluation. SEED-ML offers a resource for developing and benchmarking machine learning models, enabling research in predictive analytics, decision support systems, and computational andrology. This dataset aims to facilitate interdisciplinary collaboration between clinicians, data scientists, and AI (artificial intelligence) researchers. The dataset is publicly available in Mendeley under a CC BY 4.0 license.