Deep learning to improve the diagnostic potential of fetal brain ultrasound
PI: Namburete, Ana
Department: Computer Science (BL)
Live imaging is an essential diagnostic tool in modern medicine. Two-dimensional (2D) ultrasound is a fast, safest, and cheap tool, making it ideal to monitor the baby in the womb, and to detect any brain abnormalities during pregnancy. However, 2D images only capture 'slices' of the brain, rather than the full brain in 3D. This limits the ability to make a thorough evaluation, unless the baby is scanned using more MRI, which is more expensive and unavailable to most pregnancies. This project aims to significantly minimise the technology gap between the 2D images collected at the bedside, and the advanced brain assessment that is possible with 3D images.
In our prior work, we developed AI-based techniques to automatically reconstruct 3D images of the brain from standard clinical 2D videos. This project aims to translate our into clinical and industrial impact by developing a user-in-the-loop recommendation system that advises doctors on how and where to acquire extra images to improve the reconstructed 3D image. Our system leverages existing 2D ultrasound technology, without requiring any new equipment or substantial changes to be made to routine clinical procedures. Our solution will also broaden access to cheaper but more advanced assessment of fetal brain health. It, thus, has the potential to enhance the quality of prenatal care and contribute to better healthcare equality, worldwide.
We are partnering with the brain diagnostics company Siena Imaging (Italy) and the specialist brain centre at Leiden University Medical Centre (Netherlands). Siena Imaging has a rich track record of translating research into novel tools that provide diagnostic support to clinicians. Our clinical collaborator, LUMC, will provide expertise in ultrasound scanning of the fetal brain using 2D technology, and opportunities to scan pregnant women (>40 per day). They will also advise on the design of scanning procedures that can become part of the routine care.