Between ten and fifteen per cent of all couples in Norway are involuntarily childless. Over the past several decades, new methods for bringing about conception outside of the womb have been pursued. Referred to as assisted reproduction, only around one-third of such attempts are successful.
"Our goal is to improve the ways in which sperm and embryos are selected, thereby increasing the likelihood of achieving pregnancy and, ultimately, a successful birth," explains Trine B. Haugen, professor of biomedicine at OsloMet and the project head.
Are some sperm cells better than others?
The project team, which draws in researchers from the fields of biomedicine and information technology, plans to make use of artificial intelligence to analyse videos of sperm and embryo development during the period immediately following conception.
The overarching aim of the project is to develop better methods of selecting the sperm cells that are used to fertilise an egg—whether fertilisation occurs by means of in-vitro fertilisation or sperm is injected directly into the egg.
"We employ machine learning to locate the perfect sperm to use in assisted reproduction,” Haugen explains. “We hope to be able to predict which sperm cells are in fact capable of fertilising the egg."
Shortcomings in the current approach to sperm selection
As things stand today, the sperm cells used in assisted reproduction are selected using a microscope. Sperm are selected based on external characteristics such as mobility and appearance.
Haugen elaborates how the process works. "The sperm cells are put in a viscous drop under the microscope, at which point the person carrying out the procedure is forced to select cells somewhat at random without being able to inspect the cells closely. Typically, we select cells that are moving around,” the professor explains, “but there is no hiding the fact that this not a very thorough approach.”
We hope to be able to predict which sperm cells are in fact capable of fertilising the egg.– Trine B. Haugen
There are currently no established criteria for the kinds of sperm that are best suited for use in assisted reproduction.
"There may be important information about sperm that we currently are not aware of and hence fail to take into account when selecting," Haugen admits.
Enter artificial intelligence
The most commonly used method of assisted reproduction, in vitro fertilisation, results in failure two-thirds of the time (refer to the factbox for more on this method).
"There is enormous potential for improvement in the field of assisted reproduction,” Professor Haugen assures us.
The research team believes artificial intelligence has a potentially groundbreaking role to play.
Researchers affiliated with the project are currently bringing together data drawn from video footage of sperm cells and other kinds of data like patient information and biological factors, with the goal of obtaining new insights into sperm quality.
In search of the ideal sperm cell
“The way we do things today, when viewing sperm through a microscope in the laboratory, we look at whether sperm cells are swimming quickly or lying in a stationary position,” explains Michael A. Riegler, Chief Research Scientist at SimulaMet.
“With the help of machine learning, we are able to see how sperm cells move. This enables us to observe their pattern of movement in an entirely different way than was possible using traditional microscope observation techniques," Riegler continues.
In this way, the research team hopes to answer the question of what characteristics an ideal sperm cell possesses, and thus improve their ability to predict which cells will be able to successfully fertilise an egg.
"The main advantage of using artificial intelligence is that we can both automate sperm selection as a time-saving measure, and at the same time gain new insights into connections and patterns we haven’t previously been aware of,” the scientist explains.
Predicting how sperm cells move in real time
The project was still in its early stages at the time of writing. Nevertheless, the researchers have already made significant progress, particularly in developing new methods for predicting the quality of sperm.
The team has been busy analysing videos of sperm samples and developing algorithms intended to help improve and rationalise sperm analysis. Sperm analysis typically constitutes the first step when couples struggling to get pregnant seek medical help, and the prevailing methods have barely changed since the 1950s.
Project Manager Haugen is encouraged by the fact that the team has succeeded in predicting the movements of sperm cells in real time by employing machine learning.
“This suggests that the method could actually work."
Beginning in early 2020 the team will begin selecting the actual sperm cells to be used in assisted reproduction. The hope is that health care professionals working with assisted reproduction will be able to apply these new methods when they engage in sperm selection in their work with patients.
The research team is also preparing to embark on the next stage of the project—analysing how an embryo develops in the days immediately following fertilisation, before it is implanted in the woman’s womb.
"Our hope,” Haugen explains, “is that we will identify even better methods for selecting embryos, as this could lead to the birth of more healthy babies in Norway."
More about the project
The project Artificial intelligence: a novel tool in assisted reproduction technology is funded by the Research Council of Norway as part of the funding scheme for independent projects (FRIPRO). The project team includes researchers from the Department of Life Sciences and Health and the Department of Computer Science at OsloMet, SimulaMet and Fertilitetssenteret.
The project applies artificial intelligence (AI) methods to improve the selection of embryos and sperm for use in assisted reproduction. The primary aim of the project is to develop machine learning methods for selecting sperm cells and embryos in the early stages of the pregnancy.
The research team will also carry out a randomised study where the artificial intelligence methods they have developed are applied, as a supplement to standard treatment procedures.
Overview of the project team
Trine B. Haugen (Project Manager), Jorunn M. Andersen, Erwan Delbarre, Eldri U. Due, Mario Iliceto, Radhika Kakulavarapu, Lisbeth Charlotte Olsen and Oliwia Witczak of the Department of Life Sciences and Health, OsloMet.
Hugo L. Hammer, Akriti Sharma and Anis Yazidi of the Department of Computer Science, OsloMet.
Michael A. Riegler and Steven A. Hicks of SimulaMet, a research centre affiliated with OsloMet.
Mette H. Stensen, Anne Hancke-Framstad, Nicolai Holst, Anniken S. Nøkleby of Fertilitetssenteret.