Project description
AI for improved fetal ultrasound image solution
The ERC advanced grant PULSE project has developed innovative ultrasound imaging capabilities that have significantly improved the use of imaging technology in clinical settings. However, the lack of accurate training of sonographers in the production of high-quality diagnostic images has been identified as the major obstacle to the universal implementation of ultrasound in clinical medicine. The EU-funded PURFECT project will build on these results by developing an innovative solution supported by artificial intelligence that will guide an expert to find a standard mid-trimester fetal ultrasound imaging plane. The project will provide additional support to experts in performing more accurate scanning and contribute to the more widespread use of ultrasound in healthcare.
Objective
Perception Ultrasound for Reassuring Fetal Echo Clinical Trainees (PURFECT) builds on the results of ERC Advanced Grant PULSE to develop an AI-assistive system to guide a user to find a standard mid-trimester fetal ultrasound imaging plane. Such a system would provide reassurance to newly qualified sonographers as they begin to perform real-world diagnostic imaging, may support experts perform scanning more efficiently, and would contribute to the desired trend to use ultrasound more widely in healthcare delivery.
The novelty of the proposed system is that the guidance provided by the PURFECT solution is provided by artificial intelligence (AI) multi-modal deep learning models which predict where an expert would look. These models have been built from ultrasound video and gaze tracking of full ultrasound scans as part of the PULSE project.
The Proof of Concept award aims to build on the algorithm research conducted in PULSE to systems engineer a working PURFECT prototype of an AI-assisted ultrasound image guidance system. This will require an inter-disciplinary approach to design, build and test a system suitable for use by the intended user. In addition to addressing the engineering challenges, the project will also conduct a series of pilot usability studies to understand how AI-assisted image plane guidance can assist a trainee in diagnostic imaging and give them reassurance just as a human tutor or more experienced colleague might do. The project aims to generate licensable intellectual property and therefore will consider as well commercialization options for the developed technology for post-project follow up.
Fields of science
Keywords
Programme(s)
Funding Scheme
ERC-POC - Proof of Concept GrantHost institution
OX1 2JD Oxford
United Kingdom