Grant: Master's Award: Canada Graduate Scholarships
University of Toronto
$17,500.00
Automated facial pain detection for older adults with dementia
EF 2019-20
Résumé
"Pain occurs very frequently in older adults, but it is often undertreated. This problem is further exacerbated in older adults with dementia, who typically have more difficulty self-reporting pain due to the cognitive and linguistic impairments that are associated with dementia. Therefore, we are developing a computer vision-based tool that can monitor and automatically detect pain in older adults and send an alert to the nursing staff if a person is in pain. To codify pain, The Prkachin and Solomon Pain Intensity (PSPI) score, which ranges from 0-16, adds the intensities of 4 core action units (AUs), or facial movements, from the The Facial Action Coding System (FACS). FACS is a system that contains 44 AUs, each of which measures a certain distinguishable facial movement. There are datasets coded by certified FACS coders, and validated by self-reports, containing video frames of older adults with and without pain, that can be used to train a system to automatically detect pain. We intend on using deep learning to train a regression-based convolutional neural network (CNN) to detect pain intensity. However, like other current pain intensity estimation models, which are mostly based on supervised machine learning algorithms, we are limited by the scarce amount of FACS-labelled data. As facial expression databases are abundant online, we have obtained more than 30 non-FACS based datasets to perform unsupervised data augmentation (UDA), which utilizes both labelled and unlabelled data. In addition, since we want our pain intensity detection algorithm to be fair to a person regardless of gender, age, or dementia status, we intend to incorporate fairness with respect to these attributes into our algorithm and then study the tradeoff between fairness and accuracy."
Programme
Master's Award: Canada Graduate Scholarships
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
CIHR Funding Decisions Database
Tel que publié
Id
428992
Name
Moturu Abhishek
Region
Ontario
Country
Canada
Orgtype
Institution Paid; Research Institution
Prcname2
Special Cases - Awards Programs
Projecttitle
Automated facial pain detection for older adults with dementia
Programtype2
Studentship Programs
Orgnameinp2
University of Toronto
Cihramount2
$17,500
Cihrequipment2
$0
Programname2
Master's Award: Canada Graduate Scholarships
Keyworddelim
Artificial Intelligence; Computer Vision; Dementia; Detection; Facial Expression; Machine Learning; Neural Networks; Pain; Rehabilitation; Video
Abstract
"Pain occurs very frequently in older adults, but it is often undertreated. This problem is further exacerbated in older adults with dementia, who typically have more difficulty self-reporting pain due to the cognitive and linguistic impairments that are associated with dementia. Therefore, we are developing a computer vision-based tool that can monitor and automatically detect pain in older adults and send an alert to the nursing staff if a person is in pain. To codify pain, The Prkachin and Solomon Pain Intensity (PSPI) score, which ranges from 0-16, adds the intensities of 4 core action units (AUs), or facial movements, from the The Facial Action Coding System (FACS). FACS is a system that contains 44 AUs, each of which measures a certain distinguishable facial movement. There are datasets coded by certified FACS coders, and validated by self-reports, containing video frames of older adults with and without pain, that can be used to train a system to automatically detect pain. We intend on using deep learning to train a regression-based convolutional neural network (CNN) to detect pain intensity. However, like other current pain intensity estimation models, which are mostly based on supervised machine learning algorithms, we are limited by the scarce amount of FACS-labelled data. As facial expression databases are abundant online, we have obtained more than 30 non-FACS based datasets to perform unsupervised data augmentation (UDA), which utilizes both labelled and unlabelled data. In addition, since we want our pain intensity detection algorithm to be fair to a person regardless of gender, age, or dementia status, we intend to incorporate fairness with respect to these attributes into our algorithm and then study the tradeoff between fairness and accuracy."
Competitiondate
201912
Approvedterm2
1 yr 0 mth
Pinamesdelim
Moturu, Abhishek
Approvedterm
1 an 0 mois
Cihrcontribution2
~ $17,500~ $0~