AI Echo

CHDs (congenital heart diseases) are the most common type of birth defect, affecting nearly 1% of births per year.
About 1 in 4 babies born with a heart defect has a critical CHD.

Now that many of those algorithms don’t account for heterogeneity or variability of imaging equipment, our algorithm mainly focuses on the data from echocardiography, a ultrasound technology which captures high temporal and spatial resolutions of the heart and surrounding structure and which serves as a primary equipment for the diagnosis and follow-ups of CHD patients with the combination of rapid acquisition of image and the lack of ionizing radiation.
To date, no autonomously functioning radiology AI algorithms have gone to market; in the current market, AI still serves as a clinical decision support tool designed to assist with specific components of an imaging specialist’s workflow
While other works applying machine learning to medical images requires re-annotation by human experts, the clinical workflow for echocardiography inherently includes many measurements and calculations and is often reported through structured reporting systems; its ability to use previous annotations and interpretations from clinical reports will contribute to the building of our algorithms and accelerate the adoption of it in clinical practice.

Our deep learning models, which are supposed to rely on large volumes of annotated data to automatically learn features for subsequent detection tasks, will still have such a “small data challenge,” wherein the number of samples available for training a machine learning model is several orders lesser than the number of model parameters.