CardioAI: (Use of) AI in Fetal Cardiology

• Research type

  Research Study

• Full title

  CardioAI - (use of) AI in Fetal Cardiology : Pilot study to use machine learning to identify abnormal fetal hearts from databank of pre-existing fetal cardiac ultrasound (echocardiogram) images

• IRAS ID

  352309

• Contact name

  Umber Agarwal

• Contact email

  umber.agarwal@lwh.nhs.uk

• Sponsor organisation

  Liverpool Women's Hospital NHS Foundation Trust

• Duration of Study in the UK

  0 years, 8 months, 24 days

• Research summary

  Antenatal detection of congenital heart disease (CHD) remains limited, with <50% of cases identified during routine 18-20 week fetal anomaly screening for pregnant women, despite standardised UK Fetal Anomaly Screening Programme (FASP) guidelines. This reflects CHD heterogeneity and reliance on highly skilled obstetric
  sonographers to detect subtle abnormalities under time and workload constraints. While AI has shown potential, adoption remains limited due to poor integration of clinical expertise and limited interpretability.

  Primary Objective/Outcome
  The primary objective/outcome of this study is to collect and create a databank of pseudo-anonymised fetal cardiac ultrasound (echocardiogram) images of normal and abnormal heart structures.

  Secondary Objective/Outcome
  a) to utilise databank of pseudo-anonymised fetal cardiac ultrasound (echocardiogram) images for pilot study to develop a machine learning model to differentiate between normal and abnormal fetal hearts.
  b) to validate the Machine Learning (ML) model created with other existing data (external validation)
  c) to subsequently apply for further grants/funding to acquire larger databank of images and for further enhance (fine-tune) and validate the ML, to enable differentiation into specific types of diagnosis for the abnormal fetal hearts.

  This project has two linked phases. Phase 1 involves the creation of a pseudonymised databank of fetal cardiac ultrasound (echocardiography) images collected from routine clinical care. Phase 2 involves the exploratory development and evaluation of artificial intelligence (AI) and machine learning methods using this databank to investigate whether normal and abnormal fetal heart anatomy can be distinguished from ultrasound images. No AI model developed in this study will be used for clinical decision-making.

• REC name

  Wales REC 7

• REC reference

  26/WA/0198

• Date of REC Opinion

  15 Jul 2026

• REC opinion

  Favourable Opinion