Grant: Doctoral Research Award: Canada Graduate Scholarships
University of Toronto
$105,000.00
Machine Learning In Medicine: Evaluating Critical Barriers to Effective and Equitable Implementation Within Clinical Decision Making
EF 2022-23
Résumé
Machine learning is a novel technique that uses data to make predictions. Potential uses of machine learning within medicine are rapidly growing, including skin and radiographic image analysis. Machine learning offers much promise in clinical medicine through its ability to analyze high volumes of data and identify patterns that are not detectably by clinicians.
However, few machine learning algorithms have been shown to improve clinical outcomes, and there are important technical and practical considerations that must be addressed for the effective integration of machine learning into clinical care. My thesis will focus on studying key barriers to the effective implementation of machine learning methods in medicine.
One critical limitation of machine learning algorithms is the introduction or exacerbation of bias. My first project will be a review of all prior clinical applications of machine learning which have been found to produce bias based on sociodemographic factors, including why bias was introduced and how it could have been avoided.
A second limitation of machine learning algorithms is a lack of effectiveness in real-world use. My second project will evaluate a machine learning tool to detect delirium that is being used in clinical practice. I will evaluate both how the effectiveness of the tool changes over time in the real world, and whether the tool performs equally well in all populations (eg., based on gender, sex, ethnicity, etc.).
A third limitation of machine learning algorithms is an inability to account for clinician interactions. My third project will focus on the interaction between machine learning models and clinicians (such as trust and utilization), to develop key insights for model implementation, and strategies that allow for continued upgrading and maintenance of models after they are initially deployed.
The insights derived from these three projects will be widely applicable to machine learning solutions in medicine.
Programme
Doctoral Research Award: Canada Graduate Scholarships
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
CIHR Funding Decisions Database
Tel que publié
Id
475628
Name
Colacci Michael
Region
Ontario
Country
Canada
Orgtype
Institution Paid; Research Institution
Instname2
Health Services and Policy Research
Theme2
Clinical
Prcname2
Doctoral Research Awards - B
Projecttitle
Machine Learning In Medicine: Evaluating Critical Barriers to Effective and Equitable Implementation Within Clinical Decision Making
Programtype2
Studentship Programs
Orgnameinp2
University of Toronto
Cihramount2
$105,000
Cihrequipment2
$0
Programname2
Doctoral Research Award: Canada Graduate Scholarships
Keyworddelim
Critical Illness; In-Hospital Mortality; Inpatient Medicine; Internal Medicine; Machine Learning; Understanding And Preventing Bias In Artificial Intelligence
Abstract
Machine learning is a novel technique that uses data to make predictions. Potential uses of machine learning within medicine are rapidly growing, including skin and radiographic image analysis. Machine learning offers much promise in clinical medicine through its ability to analyze high volumes of data and identify patterns that are not detectably by clinicians.
However, few machine learning algorithms have been shown to improve clinical outcomes, and there are important technical and practical considerations that must be addressed for the effective integration of machine learning into clinical care. My thesis will focus on studying key barriers to the effective implementation of machine learning methods in medicine.
One critical limitation of machine learning algorithms is the introduction or exacerbation of bias. My first project will be a review of all prior clinical applications of machine learning which have been found to produce bias based on sociodemographic factors, including why bias was introduced and how it could have been avoided.
A second limitation of machine learning algorithms is a lack of effectiveness in real-world use. My second project will evaluate a machine learning tool to detect delirium that is being used in clinical practice. I will evaluate both how the effectiveness of the tool changes over time in the real world, and whether the tool performs equally well in all populations (eg., based on gender, sex, ethnicity, etc.).
A third limitation of machine learning algorithms is an inability to account for clinician interactions. My third project will focus on the interaction between machine learning models and clinicians (such as trust and utilization), to develop key insights for model implementation, and strategies that allow for continued upgrading and maintenance of models after they are initially deployed.
The insights derived from these three projects will be widely applicable to machine learning solutions in medicine.
Competitiondate
202211
Approvedterm2
3 yrs 0 mth
Pinamesdelim
Colacci, Michael
Approvedterm
3 ans 0 mois
Cihrcontribution2
~ $105,000~ $0~