Investigating Potential of Ethnicity Information Embedded in Medical Images
Principal Investigator
Name
Rasika Rajapakshe
Degrees
Ph.D.
Institution
BC Cancer, part of the Provincial Health Services Authority
Position Title
Senior Medical Physicist
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-1264
Initial CDAS Request Approval
Jun 17, 2024
Title
Investigating Potential of Ethnicity Information Embedded in Medical Images
Summary
This research investigates the presence of unrecognized ethnicity information in medical images using datasets from BC Cancer. The study focuses on breast and lung screening images, aiming to identify and mitigate potential biases in AI models that could lead to disparities in healthcare outcomes. By employing deep learning techniques, specifically Convolutional Neural Networks (CNNs) like ResNet and InceptionNet, the project will analyze the correlation between image features and ethnicity information. The findings will help improve the fairness and accuracy of AI-assisted diagnostic tools, ensuring consistent performance across different patient groups and enhancing overall healthcare experiences.
Aims
- Analyze the presence of unrecognized ethnicity information in BC Cancer breast and lung screening image datasets.
- Develop and fine-tune deep learning models (e.g., ResNet, InceptionNet) to predict ethnicity categories from medical images.
- Validate the models' performance using techniques like k-fold cross-validation and performance metrics such as accuracy, precision, recall, F1-score, and AUC-ROC.
- Assess the ethical implications of inferred ethnicity information on patient care, discussing potential risks and benefits.
Collaborators
Patricia Lasserre, Associate Professor, Computer Science, UBC