Hong Kong AI Male Fertility Testing with Over 96% Accuracy

The University of Hong Kong’s LKS Faculty of Medicine (HKUMed) has unveiled a pioneering artificial intelligence (AI) model that dramatically enhances the precision of male fertility assessment by accurately identifying sperm capable of fertilisation. By emulating the egg’s natural selection mechanism, this deep-learning tool promises to streamline assisted reproductive treatments (ART), reduce diagnostic variability, and improve outcomes for couples experiencing infertility.

Addressing a Global Health Challenge

Infertility affects approximately one in six couples of reproductive age worldwide, with male factors contributing to 20–70% of cases. Traditional semen analysis evaluating sperm concentration, motility, and morphology under a microscope has long been the cornerstone of male fertility diagnostics. However, its reliance on subjective visual judgement leads to significant inter-laboratory variability and limits predictive power: even men with ostensibly normal semen parameters can experience fertilisation failure in 5–25% of in vitro fertilisation (IVF) attempts.

Against this backdrop, the HKUMed research team has developed the world’s first AI model that evaluates sperm morphology based on its capacity to bind to the zona pellucida (ZP), the egg’s outer coat, thereby replicating the natural selection process of fertilisation.

The AI Model: Emulating Nature’s Screening Mechanism

Deep-Learning Framework and Validation

Leveraging advanced deep-learning techniques, HKUMed researchers trained the AI system on over 1,000 annotated sperm images, achieving an initial classification accuracy exceeding 96%. The model was subsequently validated on a vast dataset of over 40,000 sperm images drawn from 117 men diagnosed with infertility or unexplained infertility between 2022 and 2024. The results demonstrated a robust correlation between the AI-measured ZP-binding rate and actual ART outcomes, confirming the model’s clinical reliability.

Clinical Threshold and Risk Stratification

The AI tool establishes a clinical threshold of 4.9% ZP-binding sperm: men whose samples fall below this threshold are considered at heightened risk of fertility issues during IVF. By flagging patients with impaired fertilisation potential early, clinicians can tailor treatment strategies, such as adjusting insemination methods or recommending intracytoplasmic sperm injection (ICSI), to optimise success rates and minimise the emotional and financial burdens of repeated ART cycles.

Advantages over Traditional Methods

Standardisation and Reproducibility

Conventional semen analysis is labour-intensive, time-consuming, and highly dependent on the expertise of laboratory technicians. This subjectivity leads to inconsistencies across practitioners and institutions. In contrast, the AI model provides a fully automated, reproducible assessment that is independent of operator bias, ensuring consistent diagnostic criteria worldwide.

Enhanced Predictive Power

By focusing on the biological event most closely linked to fertilisation, the AI model outperforms conventional parameters in forecasting ART success. Early clinical data suggest that integrating ZP-binding analysis into fertility workups could significantly reduce the incidence of fertilisation failure, accelerating the time to pregnancy for many couples.

Efficiency and Cost-Effectiveness

Automating sperm selection analysis streamlines laboratory workflows, reducing technician workload and turnaround time. While initial deployment requires investment in imaging and computational infrastructure, the anticipated improvements in ART efficiency and reduced cycle repetition may offer long-term cost savings for fertility clinics and patients alike.

Perspectives from the Research Team

Professor William Yeung Shu-biu of HKUMed’s Department of Obstetrics & Gynaecology underscores the limitations of manual semen analysis: “This method is not only labour-intensive and time-consuming, but also highly dependent on the subjective judgement of laboratory technicians. This leads to significant variations between individuals and across laboratories, making it difficult to standardise sperm quality criteria and undermining the accuracy of male fertility evaluations.”

Future Directions and Broader Implications

The HKUMed team is currently conducting large-scale clinical trials to further validate the AI tool’s efficacy across diverse patient populations and clinical settings. Pending regulatory approvals, integration of ZP-binding analysis into standard fertility assessments could redefine best practices in andrology labs globally.

Beyond diagnostics, the technology may inspire novel sperm-selection methods for ART, potentially improving embryo quality and live birth rates. As AI continues to permeate reproductive medicine, the fusion of machine learning with physiological insights holds the promise of personalised fertility care tailored to each couple’s unique needs.

Conclusion

By harnessing AI to simulate the egg’s natural selection of sperm, HKUMed’s breakthrough model sets a new benchmark for male fertility testing. With accuracy surpassing 96% and a clear linkage to ART outcomes, this innovation paves the way for more reliable diagnoses, personalised treatment plans, and ultimately, higher success rates for couples striving to conceive. The advent of AI male fertility testing ZP-binding analysis heralds a transformative era in reproductive medicine, one where data-driven precision meets the biology of life’s very beginnings.