JBRA Assisted Reproduction 2025;29(Suppl.2 SBRA 2025):18
Poster Presentation
29^th Annual Congress of the SBRA. São Paulo/SP - Brazil, 2025
doi: 10.5935/1518-0557.20263539

P-06. An artificial intelligence (AI) model non-invasively as support to clinical decision in cycles of fertility preservation: assessment of 264 oocytes in POSEIDON groups

Camila Dutra de Souza Francisquini¹, Samara Artuso Giacomin¹, Estela Santos Oxley¹, Alessandro Schuffner¹

¹Conceber - Centro de Medicina Reprodutiva - Curitiba - PR - Brasil

Objective: To determine if AI image analysis of oocytes can identify those that have higher potential into POSEIDON groups to support medical conduct about new stimulations in fertility preservation.
Methods: This retrospective, uni-center study was conducted from June 2024 to July 2025 of patients undergoing fertility preservation. Inclusion criteria was a social fertility preservation or for accumulation of oocytes for IVF due to severe male factor or decreased ovarian reserve. Exclusion criteria was oncologic fertility preservation due to the non-attendance of anti-müllerian hormone (AMH) dosage. Non-invasive, light microscope images were taken of 264 mature oocytes post denudation and before cryopreservation. Utilizing 2-D image analysis of mature oocytes by Violet™ by Future Fertility and scoring each oocyte on a probability of blastocyst formation (%) after thawing. The patients were grouped according with POSEIDON^® criteria in four groups. In Group 1 (AMH ≥1,2ng/mL and <35 years old): 4 cycles and 63 oocytes (48 mature, 15 IVM); Group 2 (AMH ≥1,2ng/mL and ≥35 years old): 12 cycles and 123 oocytes (114 mature, 9 IVM); Group 3 (AMH <1,2ng/mL and <35 years old): 8 cycles and 39 oocytes (33 mature, 6 IVM); and Group 4 (AMH <1,2ng/mL and ≥35 years old): 8 cycles and 39 oocytes (36 mature, 3 IVM). Within each group, was separated the probability of blastocyst formation into <50% and ≥50% for mature oocytes and oocytes matured in vitro (IVM). We performed the comparison by a Two-proportions Z test (p<0.05) within the groups for these probabilities. And between the groups, was compared the percentages of blastocyst formation <50% and ≥50% for mature oocytes and for IVM oocytes.
Results: The chance of oocytes reaching ≥50% blastocyst stage after thawing is higher (p<0.05) in groups 1 (56.2%) and 2 (49.2%) (AMH groups ≥1.2ng/mL), when compared with groups 3 (27.3%) and group 4 (13.9%) (AMH groups <1.2 ng/mL). In group 1, with the best prognosis according to POSEIDON criteria, it was also observed that there is better prognosis in relation to AI for oocyte quality in potential for blastocyst formation. Only in group 1, the mature oocytes (56.2%) show chance of reaching ≥50% blastocyst stage after thawing (p<0.05) than IVM (20.0%). In group 3, with AMH <1.2 ng/mL, although the age is favorable (<35 years), there is a low probability of oocytes reaching ≥50% blastocyst after thawing (25.7%) due to the decrease in oocyte quality assessed by AI, with no difference (p>0.05) in relation to group 4 (12.8%) with the poor prognosis in POSEIDON criteria. The group 4 presented 0% chance of oocytes reaching ≥50% blastocyst stage after thawing in IVM oocytes, however, when compared with the other groups (group 1 – 20.0%; group 2 – 22.3%; group 3 – 16.7%) there was no statistical difference (p>0.05).
Conclusion: For patients with favorable age although low AMH dosage, should offer more cycles of fertility preservation due to important decrease of oocyte quality. IVM oocyte cryopreservation in poor prognosis groups can increase a number of available oocytes with good quality for the future treatment. The use of an AI system for personalized non-invasive oocyte evaluation supports a medical decision for patients regardless of age and ovarian reserve, and is especially important in cases of outliners results for decision-making for more cycles.