Using artificial intelligence to analyse ultrasound scans can detect up to 35 per cent more high-risk pregnancies than scans performed by healthcare professionals without AI decision support. This is demonstrated by results from a newly established spin-out company, Prenaital, from DTU and the University of Copenhagen, which engineers, computer scientists and doctors have set up after several years of collaboration, most recently under the auspices of the Technical University Hospital of Greater Copenhagen, TUH.
Prenaital's AI models are in a development phase, and the first product for quality assurance of ultrasound examinations is awaiting regulatory approval. The first model - an AI model for growth scanning that can detect up to 35 per cent of all foetuses at risk of abnormal growth - is expected to be on the market in 2026. The models have been developed in collaboration with sonographers, midwives and doctors at Rigshospitalet, who have identified the technology they need most, and the AI technology has then been trained at DTU using more than 10,000 images from ultrasound scans from Danish hospitals.
“Ultrasound images contain large amounts of data that the human eye cannot detect, but which can be used to identify high-risk pregnancies. This includes structures in the foetus's brain, fat percentage and tissue structures that can be used to predict the foetus's development. Today, the size and growth of the foetus are determined by measuring the head circumference, abdominal circumference and femur length on the ultrasound image, but the AI model can utilise all the information in the image,” says Professor Aasa Feragen from DTU, who is co-founder of Prenaital.
AI technology will significantly increase the benefits of the ultrasound scans that pregnant women undergo during their pregnancy, which currently only detect half of all high-risk pregnancies. In the Capital Region alone, with 22,000 pregnant women per year, 1,500 women give birth prematurely, costing society a total of DKK 800 million. Less than 20 per cent of all cases are diagnosed in time for doctors to start preventive treatment.
Detects half of all high-risk pregnancies
Co-founder of Prenaital, professor and senior physician Martin G. Tolsgaard from Rigshospitalet has been responsible for several research projects on the effect of AI support on the safety of doctors' diagnoses since 2019. Researchers at Rigshospitalet are continuing to validate the technology in a group of 200 pregnant women who are being monitored throughout their pregnancies.
"It's frustrating when we have tools that just aren't good enough, especially when we can prevent something if we have the data to make a diagnosis. I recently had a pregnant woman who came in at week 29 and went into labour with a baby that was too small to be born. She should have been detected by the ultrasound scan, but we miss over half of high-risk pregnancies. If we had only known, we could have prevented her labour and prevented a premature birth, which causes complications for the child that will follow it for the rest of its life," says Martin G. Tolsgaard.
Machine learning and neural networks
Prenaital's analysis of ultrasound scans is based on deep neural networks, which consist of many small units called neurons that are organised in layers and designed to process and analyse image data.
Before an AI-model can be used, it must be trained on a large number of e.g. ultrasound images from many different scans. Once the model has been trained, it can begin analysing new ultrasound images based on the patterns it has learned, for example to identify different parts of the foetus such as the head, heart and other organs, and compare them with normal values, thereby detecting any abnormalities at an early stage. After the analysis, AI can generate a report summarising the findings, helping doctors to make diagnoses and plan treatments.
Ready for the clinic in 2026
In 2024, Prenaital entered into an agreement on the patent rights to the AI technology and methods on which the spinout company's products are based. The company has subsequently received funding to hire four employees and has been given space at The BioInnovation Institute Foundation (BII), an international enterprise foundation with a non-profit objective. BII helps new companies in the life science sector (e.g. medicine and biotechnology) get started, and this has also been the case for Prenaital, explains CEO and founder of Prenaital Tanja Danner.
"BII's support and accelerator programme have been crucial in enabling us to establish the company, hire our first employees and lay the foundations for translating our research results into our first products, as well as hiring employees to document our work and research results. We have gained access to a unique ecosystem of knowledge, sparring and advice at BII, which has accelerated Prenatal's development so that we can get our technology out to the pregnant women and babies who are the focus of our work," says Tanja Danner.
In the coming year, Prenaital will create the processes and workflows needed to obtain approval to develop products for medical use, while also completing the first products for assessing high-risk pregnancies, which Prenaital will market in the US, the EU and Denmark from 2026. Risk models are unique and have no direct competitors in the market.
