L26/P-099 Accelerating sperm identification in non-obstructive azoospermia (NOA): A prospective comparison of AI-assisted versus manual embryologist searches in testicular tissue
Human Reproduction · 01.07.2026 · 41 (Supplement_1)
DOI: 10.1093/humrep/deag083.435 ↗Kurz gefasst
Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von M. Bailey und 10 weiteren Autor:innen.
Abstract (Originalsprache)
Study question Does artificial intelligence (AI) integration reduce the time required to identify sperm in surgically retrieved testicular tissue compared to traditional manual embryologist searches? Summary answer AI-assisted searches significantly decrease the time to first sperm detection and reduce total analysis duration for sperm-negative samples, enhancing clinical efficiency in NOA cases. What is known already Locating sperm in testicular tissue from patients with non-obstructive azoospermia (NOA) is a resource-intensive and operator-dependent process. Current protocols often necessitate extended searches lasting up to four hours to confirm the presence or absence of viable gametes. This creates a significant clinical bottleneck, particularly in high-volume laboratories where time constraints may impact the thoroughness of the search. While artificial intelligence has shown promise in other areas of embryology, its application in identifying rare sperm cells within complex testicular biopsies remains a critical but under-researched opportunity for improving laboratory workflows and patient outcomes. Study design, size, duration This prospective observational study was conducted between January and February 2025 at a private fertility clinic in the UK. Testicular tissue samples from three NOA patients were utilised to evaluate sperm detection performance. Each sample was divided into six equal parts to ensure a controlled comparison across different search methodologies. The study assessed performance under varying levels of operator experience and technological assistance to determine the impact of AI on speed and accuracy. Participants/materials, setting, methods Samples were analysed under six conditions: senior embryologist (manual vs AI-assisted), junior embryologist (manual vs AI-assisted), and AI-only. Primary outcomes included the time to identify the first sperm, the total count of sperm detected, and the total duration required to analyse an entire sample. Statistical analysis was performed using the Friedman test to compare search methods. This design allowed for the assessment of AI as both a standalone tool and a supportive aid. Main results and the role of chance The integration of AI led to a substantial reduction in search times across all experience levels. In samples where sperm were present, AI assistance significantly decreased the time to first sperm detection. For senior embryologists, search time dropped from 15.7s to 4.8s (p < 0.05), while junior embryologists saw a reduction from 24s to 9s (p < 0.05). A significant difference in performance was observed across the six methodologies tested (χ²=26.99, p < 0.0001).In cases of sperm-negative samples, the impact on laboratory efficiency was particularly pronounced. The total analysis time required to confirm the absence of sperm was 1,680 seconds with AI, compared to 3,900 seconds for manual embryologist searches. While individual operator skill influenced the manual results, the application of AI consistently enhanced speed and provided greater confidence when no sperm were identified. These findings demonstrate that AI serves as a reliable mechanism to standardise sperm searches, mitigating the variability associated with operator fatigue and subjective observation. Despite the limited sample size, the data clearly indicate that AI-assisted workflows can resolve a primary bottleneck in surgical sperm retrieval. Limitations, reasons for caution The study is limited by its small patient cohort (N = 3) and the use of samples from a single clinical site. While the performance improvements were statistically significant, further validation with larger, more diverse testicular samples is required to confirm the system’s sensitivity across various types of testicular pathology. Wider implications of the findings Automated sperm identification redefines the clinical boundaries of surgical sperm retrieval by de-coupling recovery success from laboratory time constraints. By mitigating operator fatigue and standardising rare-event detection, AI-assisted workflows facilitate exhaustive searches that were previously logistically prohibitive, potentially improving biological outcomes and expanding treatment equity for most challenging NOA cases. Trial registration number No
Zitieren
APA
Bailey, M., Popa, T., Iliev, D., Cordero, J., Rzeznik, Z., Cardenas, D., Duran-Retamal, M., Chami, A. A., Ruso, H., Ahmed-Odia, R., & Hickman, C. (2026). L26/P-099 Accelerating sperm identification in non-obstructive azoospermia (NOA): A prospective comparison of AI-assisted versus manual embryologist searches in testicular tissue. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.435
Vancouver
Bailey M, Popa T, Iliev D, Cordero J, Rzeznik Z, Cardenas D, et al. L26/P-099 Accelerating sperm identification in non-obstructive azoospermia (NOA): A prospective comparison of AI-assisted versus manual embryologist searches in testicular tissue. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.435
BibTeX
@article{Bailey2026,
author = {Bailey, M. and Popa, T. and Iliev, D. and Cordero, J. and Rzeznik, Z. and Cardenas, D. and Duran-Retamal, M. and Chami, A. Al and Ruso, H. and Ahmed-Odia, R. and Hickman, C.},
title = {L26/P-099 Accelerating sperm identification in non-obstructive azoospermia (NOA): A prospective comparison of AI-assisted versus manual embryologist searches in testicular tissue},
journal = {Human Reproduction},
year = {2026},
volume = {41},
number = {Supplement_1},
doi = {10.1093/humrep/deag083.435}
}
EUVIMED