O-012 Artificial intelligence based ultrasound for antral follicle count and ovulation triggering in ART cycles

Abstract Study question Can artificial intelligence accurately assess ultrasound images to predict low ovarian response via antral follicle count and estimate MII oocyte numbers on the triggering day? Summary answer Artificial intelligence ultrasound images evaluation predicts low ovarian response in a similar way as experimented ultrasonographers, but performs slightly lower on the ovulation triggering day. What is known already Antral follicle count (AFC) is essential for predicting ovarian response in assisted reproduction technology (ART) cycles and assessing ovarian reserve. Its accuracy depends on experienced ultrasonographers (EU), introducing variability and potential bias in clinical decisions. Follicular monitoring for ovulation triggering also requires expert interpretation, with greater variability in settings lacking skilled professionals. Artificial intelligence(AI) technology offers a standardized approach to AFC evaluation and follicular readiness assessment, ensuring consistency and accuracy. By reducing operator dependency, AI can enhance decision-making in ART cycles, particularly in resource-limited settings, improving access to reliable ovarian assessment and optimizing outcomes in assisted reproduction. Study design, size, duration Prospective cohort study. Conducted between June and December 2024 with 166 participants. Data were collected at the initial ovarian stimulation visit. Participants/materials, setting, methods We included ART patients (IVF and oocyte cryopreservation), excluding egg donors. Each underwent AFC and trigger evaluations by both an experienced ultrasonographer (EU) and an AI-based device (Folliscan™). Medical teams received only traditional ultrasound reports. Sensitivity, specificity, and AUC were calculated to predict low ovarian response (≤4 MII oocytes) with gonadotropin doses of 225-300 IU/day. Statistical analysis included t-tests, chi-square, McNemar tests, and ROC curve analysis to compare both methods. Main results and the role of chance We analyzed 166 patients (36.7 ± 3.5 years) undergoing ART, with 35.5% (59/166) exhibiting a low ovarian response (≤4 MII oocytes). AFC was lower when measured by EU than AI in both normo- (16.5 vs. 20.0, p < 0.01) and poor responders (8.6 vs. 10.0, p < 0.01). Both methods predicted poor response, with significantly lower AFC values (EU: 16.4 vs. 8.6, p < 0.01; AI: 20.0 vs. 10.0, p < 0.01). The optimal cut-off for each technique showed EU sensitivity (Se) 88.1% and specificity (Sp) 69.2% (AUC 0.87; 95% CI 0.81–0.93), while AI had Se 83.1% and Sp 68.2% (AUC 0.80; 95% CI 0.73–0.87), with no significant difference between techniques (p = 0.71). For ovulation trigger monitoring, both methods identified more large follicles in normo- than in poor responders (EU: 5.2 vs. 2.8, p < 0.01; AI: 4.1 vs. 1.6, p < 0.01), with EU detecting 1.1 more follicles than AI (p < 0.01). EU demonstrated slightly greater accuracy in predicting retrieved MII oocytes (59% vs. 66%, p < 0.01). Limitations, reasons for caution The ovulation trigger decision was based on traditional follicle count, potentially influencing results if the automated measurement had been used instead. This pilot study requires validation with a larger sample size. Wider implications of the findings AI-based ultrasound could be a valuable alternative in settings lacking experienced ultrasonographers. While not assessed for all monitoring stages, its AFC performance matched expert assessments, aiding ovarian stimulation planning. However, for ovulation triggering, it may underestimate MII oocytes, requiring further data for validation and improved accuracy. Trial registration number No