Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

University of Toronto

$59,000.00

Enhancing Accuracy and Fairness in Mobile Applications Using Wearable Physiological Sensors

EF 2023-24

Résumé
Context: Advances in mobile wearable sensing and artificial intelligence (AI) have enabled inexpensive, seamless, and continuous measurement of physiological data in naturalistic settings, catalysing an array of new health applications with tremendous promise for accelerating personalised medicine. Despite this promise, significant gaps exist between the development of AI algorithms and their ethical and meaningful implementation in clinical settings. A key barrier is the failure of AI algorithms to generalise their performance to real environments and across user populations, due to mismatches between training and testing datasets, an issue commonly seen in wearable sensor data where mobility can lead to drastic changes in user and environmental conditions. The proposed program will address this challenge by developing and evaluating approaches for mitigating data and algorithmic limitations that hinder generalizability in the context of mHealth applications employing wearable physiological sensors in real environments. Objectives: The long-term goal of my research program is to advance fair and reliable AI in the context of mobile health (mHealth) using wearable devices in naturalistic settings. My short-term goals, within the next five years, is to develop and evaluate approaches for mitigating data and algorithmic limitations that hinder generalizability in mHealth applications using wearable physiological sensors. Specifically, we will develop and evaluate algorithms to mitigate: (1) incomplete and ambiguous data labels to enhance accuracy and reduce cost; (2) data shifts resulting from variability in operating conditions in naturalistic setting to improve generalizability in heterogeneous real-world environments, and (3) data and algorithmic biases across user characterising to improve fairness. Scientific Approach: For objective 1, we will use self-supervised learning to learn the structure of unlabelled physiological data and use this representation to for classification of health states. For Objective 2, we will address data shifts using domain generalisation approaches for multi-source data. For Objective 3, we will focus on group fairness with the assumption that the sensitive attributes are known in both training and testing environments. We will address the issue of bias by (1) characterising biases in cardiac activity data collected using wearable sensors, and (2) in-process methods to increase the fairness of AI algorithms employed for classification. Significance: The proposed program will generate knowledge to address three challenges hindering the utility of wearable sensing in real world applications. The proposed work also has significant potential for commercialization as demonstrated by our existing IP and success in this area. This program will contribute to training of 10 HQP over five years, further strengthening Canada's capacity in AI.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
RGPIN-2023-05047
Name nom
Kushki, Azadeh
Department département
Biomaterials and Biomedical Engineering, Inst. of (IBBME) - Biomaterials and Biomedical Engineering, Inst. of (IBBME)
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
59000
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
2800
Research subject group en
Artificial intelligence
Research subject group fr
Artificial intelligence (computer vision, use2603)
Research subject en
Intelligence artificielle
Research subject fr
Intelligence artificielle (vision artificielle, utiliser 2603)
Application title
Enhancing Accuracy and Fairness in Mobile Applications Using Wearable Physiological Sensors
Keyword
bias; biometrics; wearable computing
Application summary
Context: Advances in mobile wearable sensing and artificial intelligence (AI) have enabled inexpensive, seamless, and continuous measurement of physiological data in naturalistic settings, catalysing an array of new health applications with tremendous promise for accelerating personalised medicine. Despite this promise, significant gaps exist between the development of AI algorithms and their ethical and meaningful implementation in clinical settings. A key barrier is the failure of AI algorithms to generalise their performance to real environments and across user populations, due to mismatches between training and testing datasets, an issue commonly seen in wearable sensor data where mobility can lead to drastic changes in user and environmental conditions. The proposed program will address this challenge by developing and evaluating approaches for mitigating data and algorithmic limitations that hinder generalizability in the context of mHealth applications employing wearable physiological sensors in real environments. Objectives: The long-term goal of my research program is to advance fair and reliable AI in the context of mobile health (mHealth) using wearable devices in naturalistic settings. My short-term goals, within the next five years, is to develop and evaluate approaches for mitigating data and algorithmic limitations that hinder generalizability in mHealth applications using wearable physiological sensors. Specifically, we will develop and evaluate algorithms to mitigate: (1) incomplete and ambiguous data labels to enhance accuracy and reduce cost; (2) data shifts resulting from variability in operating conditions in naturalistic setting to improve generalizability in heterogeneous real-world environments, and (3) data and algorithmic biases across user characterising to improve fairness. Scientific Approach: For objective 1, we will use self-supervised learning to learn the structure of unlabelled physiological data and use this representation to for classification of health states. For Objective 2, we will address data shifts using domain generalisation approaches for multi-source data. For Objective 3, we will focus on group fairness with the assumption that the sensitive attributes are known in both training and testing environments. We will address the issue of bias by (1) characterising biases in cardiac activity data collected using wearable sensors, and (2) in-process methods to increase the fairness of AI algorithms employed for classification. Significance: The proposed program will generate knowledge to address three challenges hindering the utility of wearable sensing in real world applications. The proposed work also has significant potential for commercialization as demonstrated by our existing IP and success in this area. This program will contribute to training of 10 HQP over five years, further strengthening Canada's capacity in AI.