Clinical Artificial Intelligence Fetal Echocardiography (CAIFE)
Research type
Research Study
Full title
Development of Clinical Artificial Intelligence Models in Fetal Echocardiography for the Detection of Congenital Heart Defects
IRAS ID
317510
Contact name
Aris Papageorghiou
Contact email
Sponsor organisation
University of Oxford
Duration of Study in the UK
2 years, 10 months, 15 days
Research summary
Fetal heart abnormalities (congenital heart defects) are a large, rapidly emerging global problem in child health found in approximately 0.8% to 1.2% of newborns worldwide and a leading cause of neonatal and childhood death. Approximately 25% of children born every year with different heart problems require surgery or other ways of congenital heart defect correction in the first year of life. The detection rate of fetal heart problem varies from 34% to 85%, with some countries detecting as low as 9.3% of cases before birth. If such defects are not detected before birth, these newborns are less likely to survive before or after surgery, and more likely to have developmental complications. Detection of congenital heart defects in fetal life allows early investigations of chromosomal and other fetal abnormalities, time for parents to decide on pregnancy continuation, and time for doctors to plan any fetal and newborns treatment and delivery options.
Research has shown that machine-learning models can detect fetal heart problems more rapidly and accurately than people performing the ultrasound scan. However, progress in this research field depends on the availability of large data such as ultrasound videos of normal and abnormal fetal hearts. Up to date, congenital heart defect detection based on machine-learning models is not accurate enough for practical clinical use due to the lack of ultrasound databases as congenital heart defects are rare.
In this study, we aim to develop machine-learning models that can support doctors in detecting abnormalities in fetal hearts in real-time. We plan to collect a large database consisting of ultrasound images and videos showing the anatomical structure of normal fetal hearts and those diagnosed with congenital heart defects. We will use the collected database to train the machine-learning models to characterise the anatomy of the normal fetal heart and recognise the fetal hearts with congenital heart defects. The models will be created using machine-learning techniques and will be designed to be used in ultrasound machines, including portable computers like tablets and smartphones.
REC name
East Midlands - Leicester Central Research Ethics Committee
REC reference
23/EM/0023
Date of REC Opinion
8 Mar 2023
REC opinion
Further Information Favourable Opinion