EUVIMEDEUVIMED entdecken
Crossref

L26/P-298 Holotomography-enabled, label-free 3D imaging of subcellular organisation in human oocytes: A proof-of-concept study

D.I. Iliev, C. He, T. Popa, Z. Rzeznik, K. Makropoulos, D. Souza Bittar, A. Al Chami, C. Hickman

Human Reproduction · 01.07.2026 · 41 (Supplement_1)

DOI: 10.1093/humrep/deag083.634 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von D.I. Iliev und 7 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Study question Can label-free holotomography enable non-invasive, three-dimensional characterisation of subcellular organisation in live human oocytes Summary answer We present a proof-of-concept demonstrating how holotomography combined with deep learning-based “digital dyes” enables non-invasive, three-dimensional mapping of subcellular organisation in human oocytes. What is known already Subcellular organisation represents a frontier in understanding oocyte quality and developmental competence. Mitochondrial density and spatial distribution, for instance, have been implicated in oocyte developmental potential, and may partly explain the impact of age on oocyte competence decline. Additionally, nuclear organisation may provide additional insights into an oocyte’s meiotic competence, epigenetic state and genome stability. Existing approaches to visualise these structures such as confocal and STED microscopy require dyes and phototoxic levels of illumination, rendering oocytes clinically unusable. The inability to non-invasively assess subcellular organisation limits our understanding of oocyte biology and prevents integration of organelle-level assessment into clinical practice. Study design, size, duration This was a prospective proof-of-concept study using holotomography (HT) microscopy on GV/MI oocytes donated for research between October 2025 and January 2026. The objective was to evaluate whether HT combined with a “digital dye” (an AI algorithm that predicts fluorescence patterns in unstained images based on tissue refractive index) could non-invasively map mitochondrial distribution in 3D. The specific algorithm (RI2FL) is a convolutional neural network trained on paired HT-fluorescence images of mouse embryonic fibroblasts. Participants/materials, setting, methods Oocytes were stained with MitoTracker Deep Red and imaged via both holotomography and gold-standard confocal fluorescence microscopy. The 3D volumes (308x308x180 µm) captured were aligned using image registration algorithms. Each voxel in the HT volume contained an associated RI, which was used by RI2FL digital dye to predict the probability it would fluoresce under a real-life mitochondrial dye. The predictions were evaluated using both quantitative metrics and qualitative assessment by expert embryologists. Main results and the role of chance The digital dye demonstrated reasonable quantitative accuracy and biological plausibility in predicting mitochondrial distribution. Quantitative evaluation using the Dice similarity coefficient - a measure ranging from 0 (no overlap) to 1 (perfect agreement) - yielded a score of 0.448, indicating “moderate” correspondence between the digital dye and gold-standard confocal imaging. This performance is notable given that RI2FL was trained exclusively on mouse somatic cells, not human oocytes. Qualitative assessment by embryologists confirmed the biological plausibility of the regions identified by the digital dyes: mitochondria were appropriately within the cytoplasm with no aberrant signal outside the cell or within the germinal vesicle. Additionally, the digital dye successfully captured the characteristic peripheral enrichment of mitochondria present in more mature oocytes, consistent with the existing literature. We observed pronounced inter-oocyte variability in mitochondrial organisation, with distinct spatial distributions across different oocytes. This variability was consistent between the AI predictions and fluorescence images and may carry clinical relevance, although validation in larger cohorts of clinical embryos is required. Collectively, our findings suggest that a digital dye trained specifically on human oocytes may substantially improve prediction accuracy and enhance potential clinical utility. Limitations, reasons for caution Though promising as a proof-of-concept study, validation on a larger cohort of oocytes is required, ideally with a digital dye trained on human rather than mouse cells. Furthermore, the oocytes imaged for this study were donated for research and so may not be representative of those destined for clinical use. Wider implications of the findings Our findings demonstrate for the time that digital dyes can enable non-invasive assessment of organelle distribution in human oocytes - a capability previously restricted to research settings. Beyond mitochondria, this approach may enable visualisation of other organelles and optically distinct signalling proteins, creating clinical and research opportunities without compromising usability. Trial registration number No

Zitieren

APA

Iliev, D., He, C., Popa, T., Rzeznik, Z., Makropoulos, K., Bittar, D. S., Chami, A. A., & Hickman, C. (2026). L26/P-298 Holotomography-enabled, label-free 3D imaging of subcellular organisation in human oocytes: A proof-of-concept study. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.634

Vancouver

Iliev D, He C, Popa T, Rzeznik Z, Makropoulos K, Bittar DS, et al. L26/P-298 Holotomography-enabled, label-free 3D imaging of subcellular organisation in human oocytes: A proof-of-concept study. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.634

BibTeX
@article{Iliev2026,
  author = {Iliev, D.I. and He, C. and Popa, T. and Rzeznik, Z. and Makropoulos, K. and Bittar, D. Souza and Chami, A. Al and Hickman, C.},
  title = {L26/P-298 Holotomography-enabled, label-free 3D imaging of subcellular organisation in human oocytes: A proof-of-concept study},
  journal = {Human Reproduction},
  year = {2026},
  volume = {41},
  number = {Supplement_1},
  doi = {10.1093/humrep/deag083.634}
}
Kostenlos starten →