Around 43 years ago, in vitro fertilization (IVF) ushered in modern reproductive medicine. According to the Austrian IVF register of 2019 the probability of becoming pregnant with this treatment is approximately 35 per cent. As part of a project funded by the government of Upper Austria through the business and research strategy #upperVISION2030, the fertility centre at the Kepler University Clinic is working with Software Competence Center Hagenberg (SCCH) on improving the quality assessment of blastocysts (early embryonic cell spheres) with the aid of artificial intelligence (AI) and consequently increasing the likelihood of pregnancy.

In 2019, Austrian IVF clinics performed 11,028 IVF treatments for 7,131 couples. In 9,172 cases an embryo was transferred to the uterus, resulting in 3,132 pregnancies. This treatment is both intricate and costly and also places the patients under considerable physical and emotional strain. In IVF, eggs are fertilized in a laboratory environment and the embryo is implanted in the uterus once it reaches the blastocyst stage (day 5 after fertilization). “It is important that only high-quality blastocysts are implanted,” explains Prof. Thomas Ebner, head of the fertility centre at the Kepler University Clinic. “Previously, analysis took place under a microscope, and whether or not a blastocyst was suitable was decided by consensus and a certain degree of subjectivity by a group of specialist embryologists.”

Artificial intelligence for artificial insemination

The project aims to improve the assessment of the quality of blastocysts using AI methods in such a way that the likelihood of a pregnancy following implantation is considerably increased. “This is one area where AI is particularly well suited to making a vital contribution,” says Ebner. “A variety of neural networks can be used to determine, among other things, the quality of a blastocyst, the best available blastocysts, the number of blastocysts to be implanted or the probability of a pregnancy.” These neural networks are fed and trained with individual pieces of image data, time lapse series and other clinical parameters such as the patients’ age.

Training data required

A major challenge when using AI methods is the large amounts of training data required. Particularly in medicine, generating suitable training data for classification networks is extremely labour-intensive because of the lack of patient data, the specialist knowledge of data acquisition required and much more besides. "In this project, generative adversarial networks (GANs) are being used for the first time to create synthetic image data of blastocysts and so dramatically increase the amount of data,” says Markus Manz, CEO of SCCH. “This enables us to considerably improve the networks’ training process. Because the small amount of data is particularly a problem in medicine and often limits the performance of AI applications, the results of the project could be important for many other issues as well.”

Corporate data SCCH – KUK fertility centre

The KUK (Kepler University Clinic) was formed in 2015 by the merger of Linz General Hospital, the Provincial Gynaecological and Paediatric Clinic of Upper Austria and the Wagner-Jauregg Provincial Neuropsychiatric Clinic. With more than 1800 beds and approximately 6500 employees it is the second-largest hospital in Austria. Since it was founded in 1988, the fertility centre at the Kepler University Clinic has been a focal point for the specialist field of obstetrics and gynaecology. Today, it is part of the University Clinic for Obstetrics, Gynaecology and Gynaecological Endocrinology and is headed by Prof. Peter Oppelt MBA. The centre has a special focus on research, which is reflected in the large number of publications. Among its specialities are the morphological and morphokinetic assessment of eggs and embryos and time lapse technology. www.lebenswunsch.at

Corporate data SCCH – Software Competence Center Hagenberg

Software Competence Center Hagenberg GmbH (SCCH) is an independent research centre for software in Austria and one of the associated companies of Upper Austrian Research GmbH, the lead company for research of the province of Upper Austria (member of the UAR Innovation Network). It currently has just under 100 employees. Since it was founded by the Johannes Kepler University in 1999, the COMET K1 competence centre SCCH has focused on application-oriented research in the Softwarepark Hagenberg. The main focus is on data science and software science. Its close cooperation with partners from science, especially its founding partner JKU, and with many leading companies from business and industry makes SCCH a prime example of a company with a successful strategy geared towards the entire innovation chain of education, research and business. The research focus of SCCH is on software for production as well as on data which is playing an increasingly important role owing to the advance of learning systems. Without this combination of research priorities Industry 4.0 would be unthinkable. The COMET centre Software Competence Center Hagenberg receives funding from the Ministry of Transport, Innovation and Technology (BMVIT), the Ministry of Digital and Economic Affairs (BMDW) and the government of Upper Austria under the COMET (Competence Centers for Excellent Technologies) scheme. The COMET programme is administered by the FFG (Austrian Research Promotion Agency). www.scch.at