P-085 Use of an artificial intelligence tool to assess single-sperm motility variables related to bias preference of ICSI sperm selection practice, normal fertilization, and blastocyst formation

Abstract

Abstract Study question Which single-sperm WHO motility variables are associated with ICSI sperm selection preference, normal fertilization, and blastocyst formation when assessed by an artificial intelligence platform? Summary answer Injected sperm had higher VSL, VAP, and LIN compared with non-injected. Higher STR was associated with fertilization and higher WOB and less HMP with blastulation. What is known already Sperm selection for intracytoplasmic sperm injection (ICSI) is key to its success. Currently, the procedure may be subjective although the embryologist selects rapidly from a small population of sperm. There is no consensus on specific motility variables for single-sperm selection. SiD (IVF 2.0 Limited, UK) software is able to assess all sperm in a visual field and accurately compute the motility variables for each sperm in real-time. SiDTM could help to understand the association between motility variables related to ICSI success and to assess the differences in sperm selection practice between different clinics and operators. Study design, size, duration 473 ICSI videos and their corresponding outcome were retrospectively analyzed using SiD software and assessed against normal fertilization (2PN), and blastocyst formation (BF), from four fertility clinics in different countries between July 2021 to December 2021. Ethical approval was obtained. This study was non-interventional. The study parameters were defined as sperm that were selected or non-selected, 2PN, and BF. Participants/materials, setting, methods Motility variables were computed from ICSI videos for VSL, VCL, LIN, VAP, ALH, WOB, STR as defined by the WHO manual for the examination and processing of human semen; and HMP, a metric associated with sperm-head movements. Injected sperm were manually identified and the injection was visually verified by a senior embryologist. Extracted motility variables were standardized according to each clinic and PVP concentration. Study groups were tested with a two-sided Mann-Whitney U test. Main results and the role of chance Video analysis identified 304 selected and 9543 non-selected sperm. 2PN and BF status was known for 302 and 301 injected oocytes respectively (2PN rate: 79%, BF rate: 53%). Embryologists used VSL, VAP, and LIN as the preferred motility criteria to select sperm (p < 0.001). Using data from all clinics, we found a tendency for STR (p = 0.12) and LIN (p = 0.14) for 2PN and WOB (p = 0.11) and VAP (p = 0.18) for BF. There was partial consistency on sperm selection criteria amongst clinics: Clinic1 selected VSL, VAP, and LIN as the main variables to select sperm; Clinic2 VSL, WOB, and LIN; Clinic3 VSL, VAP, and VCL; and Clinic4 ALH, VCL, and VAP (p < 0.001). For Clinic1 higher HMP showed a correlation (p = 0.06, n = 21) to 2PN, while higher WOB (p = 0.12, n = 21) did for BF. Clinic2 showed a limited association to higher STR and 2PN (p = 0.12, n = 120), and a significant association between higher WOB and BF (p < 0.05, n = 120). Clinic3’s data suggested that higher STR was associated with 2PN (p < 0.05, n = 97) and low HMP with BF (p < 0.05, n = 97). Limitations, reasons for caution This retrospective study needs to be repeated prospectively with larger numbers in more locations using embryologists with different levels of ICSI experience. Clinical outcome data such as clinical pregnancy, miscarriage, and live birth impact is still required. Wider implications of the findings SiD was able to track and extract motility features from single sperm. SiDTM could assist embryologists in real-time to select sperm during ICSI and could be used as a training tool to reduce variability among embryologists. Trial registration number RPA-2021-03