Grant: Master's Award: Canada Graduate Scholarships
University of British Columbia
$17,500.00
Master of Applied Science NSERC Application 2022
FY 2022-23
Summary
Ejection Fraction is a heart measurement that indicates how much blood gets pumped out of the left ventricle during a heartbeat. Ejection fraction is one of the main indicators of heart failure, and timely diagnosis of irregularities in ejection fraction is essential in preventing further complications. With the emergence of point-of-care ultrasound (POCUS), patients can see ultrasounds of their hearts in the comfort of their homes; however, to understand these ultrasound videos also known as echocardiograms, they still need a specialist. Recently, there have been promising improvements in the area of automating ejection fraction measurements using machine learning, and patients can receive quick feedback on the status of the health of their hearts. Such improvements can reduce the death rate and costs associated with the untimely diagnosis of heart failure. However, the main drawback of such machine learning models is their high sensitivity to factors such as the angle of measurement, zoom, etc., which are directly related to the expertise of the person doing the measurement. Another important factor is the model of the POCUS device being used. Since the POCUS device will be used by novice users, the current state-of-the-art models will face problems producing accurate results. The purpose of our research is to use the latest techniques in artificial intelligence such as contrastive learning to make the model fail-safe and generalizable to different makes and models of the POCUS device. Furthermore, our methods will focus on making the model learn and behave like a human expert and be as accurate as possible when the data is not high quality or measurements are not done using expert analysts.
Department
Program
Master's Award: Canada Graduate Scholarships
Recipient
Location
British Columbia, Canada
Recipient Type
Academia
Type
Grant
Source
CIHR Funding Decisions Database
As published
Id
486409
Name
Kondori Nima
Region
British Columbia
Country
Canada
Orgtype
Institution Paid; Research Institution
Prcname2
Special Cases - Awards Programs
Projecttitle
Master of Applied Science NSERC Application 2022
Programtype2
Studentship Programs
Orgnameinp2
University of British Columbia
Cihramount2
$17,500
Cihrequipment2
$0
Programname2
Master's Award: Canada Graduate Scholarships
Keyworddelim
Artificial Intelligence; Contrastive Learning; Deep Learning; Ejection Fraction Estimation; Graph Neural Networks; Heart; Heart Failure; Ischemic Stroke
Abstract
Ejection Fraction is a heart measurement that indicates how much blood gets pumped out of the left ventricle during a heartbeat. Ejection fraction is one of the main indicators of heart failure, and timely diagnosis of irregularities in ejection fraction is essential in preventing further complications. With the emergence of point-of-care ultrasound (POCUS), patients can see ultrasounds of their hearts in the comfort of their homes; however, to understand these ultrasound videos also known as echocardiograms, they still need a specialist. Recently, there have been promising improvements in the area of automating ejection fraction measurements using machine learning, and patients can receive quick feedback on the status of the health of their hearts. Such improvements can reduce the death rate and costs associated with the untimely diagnosis of heart failure. However, the main drawback of such machine learning models is their high sensitivity to factors such as the angle of measurement, zoom, etc., which are directly related to the expertise of the person doing the measurement. Another important factor is the model of the POCUS device being used. Since the POCUS device will be used by novice users, the current state-of-the-art models will face problems producing accurate results. The purpose of our research is to use the latest techniques in artificial intelligence such as contrastive learning to make the model fail-safe and generalizable to different makes and models of the POCUS device. Furthermore, our methods will focus on making the model learn and behave like a human expert and be as accurate as possible when the data is not high quality or measurements are not done using expert analysts.
Competitiondate
202212
Approvedterm2
1 yr 0 mth
Pinamesdelim
Kondori, Nima
Approvedterm
1 an 0 mois
Cihrcontribution2
~ $17,500~ $0~