EUVIMEDEUVIMED entdecken
Crossref

L26/P-086 Building the foundations for AI-based sperm prioritisation strategies: hyperspectral sperm fingerprints reveal differences between embryologist-selected and unselected sperm within the same sample

N. Garrido Puchalt, D. Muñoz-Pellicer, R. Trinidad-Lahoz, P.R. Rosa María, M. Gil-Juliá, D.L.S. Jose María, M. Laura, R. Rivera-Egea, M.J. De los Santos

Human Reproduction · 01.07.2026 · 41 (Supplement_1)

DOI: 10.1093/humrep/deag083.422 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von N. Garrido Puchalt und 8 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Study question Do hyperspectral profiles of different spermatozoa within the same sample differ between morphologically and motility-competent sperm selected by embryologists and unselected sperm? Summary answer Hyperspectral sperm fingerprints revealed statistically significant differences between embryologist-selected and unselected sperm within the same sample, highlighting biochemical features that may guide AI-assisted sperm selection. What is known already Sperm selection in assisted reproduction is mainly based on morphology and motility, requiring extensive embryologist expertise and being prone to subjectivity and inter-operator variability. Non-invasive techniques such as hyperspectral imaging (HS) enables the acquisition of single-cell biochemical fingerprints, providing a promising and ICSI-compatible framework for objective and reproducible sperm selection. Our group previously reported promising results using sperm HS-based algorithms to predict fertilization and embryo development outcomes. However, no studies have investigated the biological information encoded by hyperspectral sperm fingerprints. Consequently, it is unclear whether these profiles differ between ICSI-apt spermatozoa and the unselected sperm population within the same sample. Study design, size, duration Prospective study including 10 patients undergoing ICSI cycles using autologous ejaculate sperm. For each patient, the remaining sperm sample from the procedure was used to acquire hyperspectral images of up to 75 sperm cells: up to 50 were randomly selected from the sample, without applying any selection criteria, and up to 25 were selected by the embryologist following standard clinical practice. A total of 667 hyperspectral images were acquired. Data collection performed in October 2025 Participants/materials, setting, methods For each patient, sperm cells were obtained from the PVP drop and immobilized in a clean droplet using the remaining ICSI dishes. Images were acquired using a 4200M Microscope coupled to an IX73 inverted microscope with micromanipulator systems and TruScope v.1.17 software. Embryologist-selected versus non-selected sperm spectra were compared using wavelength-wise Welch tests (≥10 contiguous wavelengths). Intra-group heterogeneity within selected sperm was quantified using pairwise spectral distances and summarized per patient. Main results and the role of chance Wavelength-wise Welch tests identified patient-dependent spectral regions differentiating embryologist-selected from non-selected sperm. Applying a robustness criterion based on contiguous bands (≥10 consecutive wavelengths), at least one statistically significant spectral region was observed in 6 of 10 patients, indicating within-patient differences between selected and non-selected sperm. These results suggest that sperm deemed morphologically and motility-competent by the embryologist exhibit distinct hyperspectral profiles compared with the remaining sperm population within the same sample. Beyond these between-group differences, intra-group heterogeneity was quantified within the embryologist-selected subset using pairwise spectral distances among all selected sperm. Summary statistics of these distributions revealed that most selected sperm were spectrally homogeneous, while a limited number of highly distinct fingerprints drove the upper tail of dissimilarity in some patients. This structured heterogeneity highlights subtle biochemical and structural differences not detectable by visual assessment alone. Together, these results demonstrate that hyperspectral imaging captures biologically meaningful variability both between and within embryologist-selected sperm populations. Rather than replacing clinical expertise, HS provides a non-invasive molecular readout that can support downstream AI-based stratification and ranking of spermatozoa within clinically viable pools, extending selection beyond the intrinsic limits of naked-eye evaluation. Limitations, reasons for caution This proof-of-concept study is limited by the number of patients and the short acquisition window. Given the large number of wavelength-wise comparisons, this analysis should be interpreted primarily as identifying relevant spectral regions rather than providing definitive statistical inference. Wider implications of the findings Hyperspectral sperm fingerprints provide a compact, non-invasive representation of biochemical and structural differences underlying embryologist selection. By revealing both selected–unselected contrasts and intra-sample heterogeneity, this work lays the foundations for AI-based sperm prioritisation strategies and motivates future development of objective decision-support tools linked to clinically relevant outcomes. Trial registration number No

Zitieren

APA

Puchalt, N. G., Muñoz-Pellicer, D., Trinidad-Lahoz, R., María, P. R., Gil-Juliá, M., María, D. J., Laura, M., Rivera-Egea, R., & Santos, M. D. L. (2026). L26/P-086 Building the foundations for AI-based sperm prioritisation strategies: hyperspectral sperm fingerprints reveal differences between embryologist-selected and unselected sperm within the same sample. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.422

Vancouver

Puchalt NG, Muñoz-Pellicer D, Trinidad-Lahoz R, María PR, Gil-Juliá M, María DJ, et al. L26/P-086 Building the foundations for AI-based sperm prioritisation strategies: hyperspectral sperm fingerprints reveal differences between embryologist-selected and unselected sperm within the same sample. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.422

BibTeX
@article{Puchalt2026,
  author = {Puchalt, N. Garrido and Muñoz-Pellicer, D. and Trinidad-Lahoz, R. and María, P.R. Rosa and Gil-Juliá, M. and María, D.L.S. Jose and Laura, M. and Rivera-Egea, R. and Santos, M.J. De los},
  title = {L26/P-086 Building the foundations for AI-based sperm prioritisation strategies: hyperspectral sperm fingerprints reveal differences between embryologist-selected and unselected sperm within the same sample},
  journal = {Human Reproduction},
  year = {2026},
  volume = {41},
  number = {Supplement_1},
  doi = {10.1093/humrep/deag083.422}
}
Kostenlos starten →