Grant: Discovery Grants Program - Individual
Université de Sherbrooke
$41,000.00
Fully-automatic, self-corrective, and explainable echocardiographic image analysis
EF 2023-24
Résumé
The goal of this proposal is to develop and thoroughly validate highly effective US cardiac image analysis solutions while at the same time, tackling the following four challenges 1) be capable to combine heterogeneous data (image, age, body mass index, ejection fraction, etc.); 2) be robust against age and sex biases; 3) be interpretable; 4) be capable of efficiently estimate and validate local and regional strain from B-mode US images. To do so, we ought to reach these three objectives: OBJECTIVE 1. Build a massively large (~100,000 subjects) cardiac B-mode US dataset with gold standard annotation based on raw unlabeled data. While this large dataset brings great benefits, it nonetheless carries important challenges that calls for fundamental work. OBJECTIVE 2. Develop an efficient local and regional strain estimation method. To do so, we shall adapt the dataset developed in Obj.1 to make it possible to train deep speckle tracking methods that are squarely robust to speckle decorrelation, thanks to an highly efficient ultrasound physic-based simulator called SIMUS. This synthetic dataset will make it possible to train bias-free local and regional strain estimators. OBJECTIVE 3. Multi-channel and interpretable cardiac disease assessment. the goal is to leverage the models and data from Obj.1 and 2 in order to propose interpretable diagnostic models that combine heterogeneous data. Also, since part of the datasets will be annotated by several experts, the propose models shall output inter-expert variability alongside of their predictions.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2023-04584
Name nom
Jodoin, Pierre-Marc
Department département
Informatique
Organization id
68
Institution établissement
Université de Sherbrooke
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
41000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1102
Area of application group en
Health, education and social services
Area of application group fr
Santé publique, éducation et services sociaux
Area of application en
Biomedical engineering
Area of application fr
Génie biomédical
Research subject code
2805
Research subject group en
Artificial intelligence
Research subject group fr
Learning and inference theories
Research subject en
Intelligence artificielle
Research subject fr
Théories de l'apprentissage et l'inférence
Application title
Fully-automatic, self-corrective, and explainable echocardiographic image analysis
Keyword
Cardiac imaging; Cross-attention networks; Deep learning; Echocardiography; Explainability; Image segmentation; Medical imaging; Motion tracking; Uncertainty estimation
Application summary
The goal of this proposal is to develop and thoroughly validate highly effective US cardiac image analysis solutions while at the same time, tackling the following four challenges 1) be capable to combine heterogeneous data (image, age, body mass index, ejection fraction, etc.); 2) be robust against age and sex biases; 3) be interpretable; 4) be capable of efficiently estimate and validate local and regional strain from B-mode US images. To do so, we ought to reach these three objectives: OBJECTIVE 1. Build a massively large (~100,000 subjects) cardiac B-mode US dataset with gold standard annotation based on raw unlabeled data. While this large dataset brings great benefits, it nonetheless carries important challenges that calls for fundamental work. OBJECTIVE 2. Develop an efficient local and regional strain estimation method. To do so, we shall adapt the dataset developed in Obj.1 to make it possible to train deep speckle tracking methods that are squarely robust to speckle decorrelation, thanks to an highly efficient ultrasound physic-based simulator called SIMUS. This synthetic dataset will make it possible to train bias-free local and regional strain estimators. OBJECTIVE 3. Multi-channel and interpretable cardiac disease assessment. the goal is to leverage the models and data from Obj.1 and 2 in order to propose interpretable diagnostic models that combine heterogeneous data. Also, since part of the datasets will be annotated by several experts, the propose models shall output inter-expert variability alongside of their predictions.