Echocardiographic and Clinical Health Observations for AI Diagnostics (ECHO-AID)

Research summary

Assessing cardiac ultrasound imaging (echocardiograms) requires a high level of skill and training for accurate and effective decision-making. The use of artificial intelligence in image analysis has increased in popularity in recent years and aims to reduce human error, reduce variability, and speed up image processing time. We, and others, have developed and have access to various artificial intelligence algorithms and platforms that make use of echocardiogram images and data from these scans to then construct diagnoses or provide severity levels of disease. A key step to move towards potential clinical use of these AI tools is evaluation of their utility, accuracy and acceptability of application across a wide variety of different clinical, demographic and patient characteristics. This study therefore aims to collect transthoracic echocardiogram data from a range of centres across the UK that can be used to test current and future developed models. This study will gather data retrospectively from patients who have received an echocardiogram scan for any reason. We will collect relevant patient demographics and health data as well as data and results from their echocardiogram scan including metrics and diagnoses. In addition, we will collect details of, reason for echocardiogram, clinic or echocardiogram scan slot appointment type, method or route of referral to echocardiogram scan, and the clinical reason for the echocardiogram scan. Where further information is available at the time of data collection on the patient clinical pathway for those having an echocardiogram at each site such as surgical procedures or health outcomes relevant to the echocardiogram this will also be collected The data will be used to test accuracy of algorithms developed to quantify or diagnose disease as well as determine the current manner in which patients with different conditions are being seen and diagnosed to better understand the optimal time for application of different AI tools. This information will help identify best ways to implement artificial intelligence models into clinical practice in the future.

Principal Investigator

Prof Paul Leeson

Contact us

Email: oxfordheartresearch@ouh.nhs.uk

IRAS number

351780