New AI technology could improve the speed and accuracy of live sperm analysis, assisting infertility diagnosis and selection of sperm for ICSI.
ICSI is a variant of IVF where a single sperm is selected and injected into the egg, rather than the sperm and egg being mixed in a dish to encourage fertilisation to occur. Researchers at Monash University and Monash IVF in Melbourne, Australia, collaborated to create an AI model to assist practitioners in identifying the best quality sperm from live samples to use in ICSI.
'It can sometimes take hours to sift through a sample to find the best sperm for injection,' said Dr Deirdre Zander-Fox, Monash IVF chief scientific officer. 'We believe AI can make the process much faster and give patients improved outcomes, while still allowing our highly trained embryologists to have oversight of the process.'
Computer-assisted methods of grading sperm already exist, but many look only at the head of the sperm. Publishing their results in Advanced Intelligent Systems, the researchers developed a model trained on high-resolution images that analysed the shape and structure of the head, midpiece and tail of the sperm to determine overall quality. The results were compared to assessments by three andrology experts.
They found that the model could analyse images with 93 percent accuracy in under one second, and could maintain a high level of accuracy in lower-resolution images.
'The consistency and reliability of our AI model provides unprecedented accuracy in live sperm morphology classification,' said lead researcher Dr Reza Nosrati who specialises in mechanical engineering. 'By providing a clear and precise analysis of sperm quality, it offers promising opportunities for enhancing clinical sperm selection practices and reducing day-to-day variability in clinics.'
'Following on from this research, Monash IVF hopes to create an AI algorithm that could be used to power a sperm selection device to guide our embryologist on which sperm to choose,' said Dr Zander-Fox.
Leave a Reply
You must be logged in to post a comment.