Grant: DRA : Canada Graduate Research Scholarship - Doctoral Program (CGRS D)
University of Ottawa
$120,000.00
Advancing Prenatal Care: Machine Learning-Powered Detection of Congenital Heart Disease from Ultrasound Scans
EF 2025-26
Résumé
Congenital heart disease (CHD) is the most common birth defect and a leading cause of newborn death and lifelong disability in Canada and worldwide. While early detection of CHD through prenatal ultrasound can be life-saving, 20-50% of cases still go undiagnosed before birth. Each missed diagnosis represents a lost opportunity to save lives or improve quality of life.
We propose to use an artificial intelligence (AI) model to advance our ability to detect the presence of CHD on routine prenatal ultrasound scans. Most frequently AI is used to identify patterns in still images. However, our approach uses both still images and video clips from the ultrasound. This will enhance the AI model with valuable information, such as heartbeat patterns, and facilitate its integration into live ultrasound scans. To our knowledge, no published work has yet developed video-based CHD detection models at scale.
We will build two AI models for this task. The first will identify the presence of CHD in a fetus. This will help ensure the mother delivers at a hospital with the right equipment and expertise needed for immediate care. The second AI model will determine which CHD subtype it is. This helps physicians anticipate the type of immediate care the baby will require, such as surgery, and provides families with the information needed to make informed decisions about the pregnancy, delivery and care after birth.
To train these AI models we are leveraging a large dataset of prenatal ultrasound scans from patients who received prenatal care through The Ottawa Hospital and CHEO between 2009-2024. The model training techniques we use will allow our model to perform reliably across different hospitals and patient populations.
To support deployment into clinical practice, we will also integrate explainability into our models to enable a clearer understanding of their decision-making processes, helping ensure there is no bias and enhancing trust in our model.
Programme
DRA : Canada Graduate Research Scholarship - Doctoral Program (CGRS D)
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
CIHR Funding Decisions Database
Tel que publié
Id
559053
Name
Erman Aylin
Region
Ontario
Country
Canada
Orgtype
Institution Paid; Research Institution
Instname2
Circulatory and Respiratory Health
Theme2
Clinical
Prcname2
DT1 (Biomedical and/or Clinical Research) CGRS Doctoral
Projecttitle
Advancing Prenatal Care: Machine Learning-Powered Detection of Congenital Heart Disease from Ultrasound Scans
Programtype2
Studentship Programs
Orgnameinp2
University of Ottawa
Cihramount2
$120,000
Cihrequipment2
$0
Programname2
DRA : Canada Graduate Research Scholarship - Doctoral Program (CGRS D)
Keyworddelim
Artificial Intelligence; Congenital Heart Disease; Machine Learning; Obstetrics
Abstract
Congenital heart disease (CHD) is the most common birth defect and a leading cause of newborn death and lifelong disability in Canada and worldwide. While early detection of CHD through prenatal ultrasound can be life-saving, 20-50% of cases still go undiagnosed before birth. Each missed diagnosis represents a lost opportunity to save lives or improve quality of life.
We propose to use an artificial intelligence (AI) model to advance our ability to detect the presence of CHD on routine prenatal ultrasound scans. Most frequently AI is used to identify patterns in still images. However, our approach uses both still images and video clips from the ultrasound. This will enhance the AI model with valuable information, such as heartbeat patterns, and facilitate its integration into live ultrasound scans. To our knowledge, no published work has yet developed video-based CHD detection models at scale.
We will build two AI models for this task. The first will identify the presence of CHD in a fetus. This will help ensure the mother delivers at a hospital with the right equipment and expertise needed for immediate care. The second AI model will determine which CHD subtype it is. This helps physicians anticipate the type of immediate care the baby will require, such as surgery, and provides families with the information needed to make informed decisions about the pregnancy, delivery and care after birth.
To train these AI models we are leveraging a large dataset of prenatal ultrasound scans from patients who received prenatal care through The Ottawa Hospital and CHEO between 2009-2024. The model training techniques we use will allow our model to perform reliably across different hospitals and patient populations.
To support deployment into clinical practice, we will also integrate explainability into our models to enable a clearer understanding of their decision-making processes, helping ensure there is no bias and enhancing trust in our model.
Competitiondate
202511
Approvedterm2
3 yrs 0 mth
Pinamesdelim
Erman, Aylin
Approvedterm
3 ans 0 mois
Cihrcontribution2
~ $120,000~ $0~