UZ Brussel develops AI tool for faster and more efficient detection of sperm cells
Researchers from Brussels IVF, the centre for reproductive medicine at UZ Brussel, have developed an innovative AI-based system together with Robovision Healthcare that can detect sperm cells much faster and more accurately following a testicular sperm extraction (TESE). This procedure is used in men whose ejaculate contains no sperm cells. The new technology, named T’easy, could mark a significant step forward in fertility treatments. The study results were recently presented at the annual congress of the European Society of Human Reproduction and Embryology (ESHRE) in Paris.
When no sperm cells are found in a man’s ejaculate, a TESE procedure (Testicular Sperm Extraction) can be performed. This involves removing a small tissue sample from the testes to isolate sperm cells for use in ICSI—a form of IVF where a single sperm cell is injected directly into an oocyte. TESE gives men with severely reduced fertility the possibility of having genetically own children.
The technique was first applied at UZ Brussel in 1995 for men with impaired sperm production in the testicles. Since then, this combined strategy has become a routine procedure worldwide—though success rates still vary depending on whether or not sperm cells can be retrieved from the tissue.
Faster and more accurate detection with AI
“With T’easy, we can detect sperm cells more quickly and more precisely, giving couples faster clarity about the likelihood of success of their fertility treatment. It also reduces the risk of missing usable sperm cells. In extreme cases, the search for sperm can take more than 15 hours. The expression ‘needle in a haystack’ is very much applicable here. At the same time, the tool also provides the information needed to make an informed decision to stop searching when no quality sperm cells are found,” says Koen Wouters, clinical embryologist at Brussels IVF.
For care workers, the application represents significant time savings and allows them to focus their expertise where it is most needed. Manually detecting sperm cells in testicular tissue is labour-intensive, time-consuming and requires specialised staff, as sperm cells in such samples are often scarce and hard to find.
Promising results
T’easy combines an app, a specially adapted microscope, and artificial intelligence to automatically and rapidly detect sperm cells in testicular tissue. The app controls the microscope, ensures sharp image capture, analyses the images with AI, and presents the results in a clear, user-friendly interface. The tool was developed in collaboration with Robovision and with the support of Vlaio and Innoviris. For now, T’easy is only being used within a research setting.
Development began in January 2022 with training of the AI model using carefully labelled data. A total of 5,373 images from 22 TESE samples were captured and double-annotated by experienced staff. A first pilot study followed in November 2024, in which the system was successfully tested.
The research team trained the AI using more than 13,000 labelled sperm cells across over 5,000 images collected between January 2022 and November 2024. In initial testing, the system achieved high accuracy, with a recall (correctly identified sperm cells) of 95% and a precision (proportion of identified cells that were actually sperm) of 94.8%.
In the pilot study, T’easy detected an average of 98% of present sperm cells and completed the analysis in just 10 minutes—more than twice as fast as experienced lab technicians, who took an average of 24 minutes.
Further refinement
Although the results are promising, challenges remain. The researchers point to issues with image quality and highlight the need for more diverse datasets to further improve the model’s robustness.
Prof. Dr. Herman Tournaye, head of Brussels IVF, explains: “We hope that, in the long run, T’easy can further increase the chances of achieving pregnancy. Future research will help refine the technique and extend its application to other types of sperm samples, such as ejaculates with extremely low sperm counts or to ensure that no sperm cells remain present following a vasectomy.”
Brussels IVF and Robovision Healthcare are currently seeking a (commercial) partner to support further research and make the broad clinical application of this technology in fertility treatment possible.
Karolien De Prez