+237 677 439 328 info@hrfbuea.org UB Junction, Buea, Cameroon

About CHD AI


A congenital heart defect is a problem with the structure of the heart that a child is born with. A congenital heart defect (CHD) results when the heart, or blood vessels near the heart, don’t develop normally before birth. Such defects result when a mishap occurs during heart development soon after conception — often before the mother is aware that she is pregnant. The word “congenital” means existing at birth Some congenital heart defects in children are simple and don't need treatment. Other congenital heart defects in children are more complex and may require several surgeries performed over a period of several years. Learning about your child's congenital heart defect can help you understand the condition and know what you can expect in the coming months and years. Congenital heart defects can change the way the heart pumps blood. They may make blood flow too slowly, go the wrong way, or block it completely.

Project Objective

This project seeks to design an Al model to automatically extract optimal images/video clips from postnatal echocardiogram of neonates (0-28 days) for onward transmission to experts. The new technology will allow low skilled sonographers serving local birthing centres in sub-Saharan Africa (SSA) to conduct an echocardiogram and extract and transmit accurate images to a remote expert for interpretation and diagnosis of congenital heart defect (CHD). The technology will:
Reduce the workload for the limited available experts in SSA who are having to deal with more cases needing a diagnosis confirmation as uptake of postnatal screening of CHD by pulse oximetry increases
Reduce costs and the need for patients to travel thousands of kilometers for diagnosis confirmation at expert facilities located mainly in urban cities, and
Increase the rate of CHD detection and early diagnosis, allowing for earlier treatment.

In the future this technology could evolve to facilitate prenatal echocardiography, diagnosis or classification of risk score for a particular CHD subtype, and predicting treatment options and survival. The ultimate goal is to deploy the technology across SSA and other low resource settings to reduce workload on the few available experts and improve the detection rate, accuracy of CHD diagnosis and ultimately care of affected infants. In the current proposal, we specifically aim to:

  • to develop and evaluate the performance of an Al model that will automatically extract and label optimal images/video clips during a postnatal echocardiography scan,
  • to explore use of automated real-time feedback regarding cardiac plane view and scan completeness,
  • to develop standard operating procedures and instructional videos for non-experts on how to manoeuvre the probe during an echocardiography scan to obtain optimal video stream of the heart
  • to investigate the diagnostic utility of the images/video clips produced from the Al model in relation to their use in clinical practice, and
  • to explore challenges of integrating the Al technology developed in clinical practice, including Al awareness training needs