EUVIMEDEUVIMED entdecken
Crossref

L26/P-784 Automated assessment of follicle number per ovary for standardized detection of polycystic ovarian morphology

P. Wygocki, A. Zapała, G. Dolanowski, M. Ulfig, M. Zieleń, M. Wrochna, K. Zieliński, N. Gajewska, D. Drzyzga, M. Brzozowski

Human Reproduction · 01.07.2026 · 41 (Supplement_1)

DOI: 10.1093/humrep/deag083.1111 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von P. Wygocki und 9 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Study question Does AI-assisted automated follicle counting improve standardized and sensitive identification of patients fulfilling the FNPO ≥20 criterion for PCOM? Summary answer The AI system demonstrated significantly higher sensitivity in detecting patients fulfilling the FNPO ≥20 PCOM criterion compared with routine clinical assessment. What is known already Polycystic ovarian morphology (PCOM) is one of the diagnostic features of polycystic ovary syndrome (PCOS) and is defined in recent ESHRE guidelines as follicle number per ovary (FNPO) ≥20 in at least one ovary in adults. Although characteristic ultrasound patterns may be visually apparent, objective follicle quantification is required to confirm whether this threshold is fulfilled. In routine practice, systematic counting is often limited by time constraints and inter-operator variability, leading to inconsistent classification, particularly in borderline cases. As ovarian morphology represents only one component of PCOS diagnosis, high-precision and standardized ultrasound assessment is essential. Study design, size, duration This retrospective validation study evaluated an independent test set of 212 two-dimensional ovarian ultrasound cine-loops from 79 patients collected across five fertility clinics in Poland between 2019 and 2021. Participants/materials, setting, methods Ground truth follicle counts were established by manual annotation with expert sonographer validation. An AI-based deep learning system (FOLLISCAN) automatically quantified follicles from two-dimensional ultrasound recordings. The independent test set was randomly selected and excluded from model training. The study population had a mean age of 31.8 ± 6.1 years and mean AMH of 3.63 ± 2.31 ng/mL, representing a wide ovarian reserve spectrum. AI-based FNPO ≥20 classification was compared with routine clinical assessments. Main results and the role of chance Based on ground truth evaluation, for 33 of 212 cine-loops the FNPO ≥20 criterion for PCOM was met. Both the deep learning model and archival reports demonstrated high precision for PCOM detection (96% and 93%, respectively). However, a marked difference was observed in recall (sensitivity), with the model achieving 79% compared with 39% for routine clinical assessment. The improvement in detection performance was statistically significant, as confirmed by the exact McNemar test for paired observations (p = 0.00097). The greatest clinical benefit was observed among patients with follicle counts near the diagnostic cut-off, representing a diagnostic grey zone where visual assessment alone often failed to reliably determine whether formal criteria were fulfilled. These findings indicate that although overt polycystic ovarian morphology may be readily recognized by clinicians, standardized automated follicle counting provides critical objective support in accurately classifying patients according to updated diagnostic thresholds. Limitations, reasons for caution The study is limited by its retrospective design and relatively small sample size. Only two-dimensional baseline ultrasound recordings were analyzed, and prospective multicenter validation is required. Wider implications of the findings High-sensitivity automated follicle counting supports standardized ultrasound assessment recommended by international guidelines and may reduce misclassification in borderline cases. This approach could enhance diagnostic consistency and inform future PCOS diagnostic pathways, including adolescent populations where PCOM is a mandatory criterion. Trial registration number No

Zitieren

APA

Wygocki, P., Zapała, A., Dolanowski, G., Ulfig, M., Zieleń, M., Wrochna, M., Zieliński, K., Gajewska, N., Drzyzga, D., & Brzozowski, M. (2026). L26/P-784 Automated assessment of follicle number per ovary for standardized detection of polycystic ovarian morphology. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.1111

Vancouver

Wygocki P, Zapała A, Dolanowski G, Ulfig M, Zieleń M, Wrochna M, et al. L26/P-784 Automated assessment of follicle number per ovary for standardized detection of polycystic ovarian morphology. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.1111

BibTeX
@article{Wygocki2026,
  author = {Wygocki, P. and Zapała, A. and Dolanowski, G. and Ulfig, M. and Zieleń, M. and Wrochna, M. and Zieliński, K. and Gajewska, N. and Drzyzga, D. and Brzozowski, M.},
  title = {L26/P-784 Automated assessment of follicle number per ovary for standardized detection of polycystic ovarian morphology},
  journal = {Human Reproduction},
  year = {2026},
  volume = {41},
  number = {Supplement_1},
  doi = {10.1093/humrep/deag083.1111}
}
Kostenlos starten →