P-087 Expanding the comparison of an AI automated semen analyses for concentration determinations with manual analyses and a cell counting system

Abstract

Abstract Study question Can comparable results be obtained for sperm concentrations determinations in differing clinical settings when using an AI driven analyser, a cytometer, and WHO manual counts? Summary answer Outcomes suggest that comparable results can be obtained when utilising the differing counting techniques. What is known already Despite robust methods described by WHO, protocol variations exist, creating a lack of reproducible and consistent semen analysis results in standard laboratories, as described by Bjorndahl etal 2016. Human error can influence results where WHO protocols are not utilised in conjunction with rigorous IQA and EQA. A human performing an analysis is considered the gold standard, but it's apparent, as demonstrated by time-lapse embryo evaluation, that human assessments can be enhanced with machine assistance to increase accuracy. To establish the benefit of machine assistance, it is necessary to demonstrate comparable results with the manual operator and reliable machine cell counters. Study design, size, duration To assess whether comparable results could be obtained with a well-trained reproductive scientist and a cell counting system, two reputable IVF clinics were selected to perform their standard analyses and perform a repeat test with an automated analyser. A number of semen analyses were performed over a six-month time-period for patients attending IVF clinics for a pre-treatment evaluation. Participants/materials, setting, methods 107 samples were collected and analysed between June 2022 to December 2023 in laboratories in Copenhagen and London. Participants were selected on a random basis. The semen analyses were performed in a manner described by the WHO6, with an additional assessment with a NucleoCounter™ chemometec in the Danish laboratory. Simultaneously, a concentration determination was performed on a Mojo AISA ™. (Artificial Intelligence Semen Analysis system). Main results and the role of chance Manual concentration assessments, the cytometer data and the AISA concentration results were plotted and examined. Comparable values were obtained when data for the manual assessment v AI machine assessment and the data obtained for fluorescent marked sperm v AI analysis were assessed. A significant Spearman correlation was obtained by the manual analysis team and the team who used the cytometer counting system, values of 0.93 and 0.99, (error average 17.5 and 24.4) respectively. Limitations, reasons for caution Increasing the sample frame of this current study, including more laboratories for a multi-centre trial would increase the power of the findings. Inter and intra operator and repeat machine evaluations are planned to establish the human variations during analysis and establish if this can be mitigated by machine assistance. Wider implications of the findings The results demonstrate that a machine assisted semen analysis is comparable to a gold standard semen analysis performed by a reputable laboratory staff either when compared to a manual operator and to a cell sorting system where sperm are labelled with a fluorescent probe. Trial registration number N/A