Graduate Thesis Or Dissertation
 

Machine Learning for Precision Medicine: Application to Cancer Chemotherapy Response Prediction and Wound Healing Status Assessment

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/f7623k898

Descriptions

Attribute NameValues
Creator
Abstract
  • Ph.D. candidate Qi Wei's thesis consists of two projects: Chemotherapy Project: a study based on the research paper "Predicting chemotherapy response of various cancer types using a variational auto-encoder approach" submitted to the bioRxiv preprint archive and accepted by the BMC bioinformatics; and Wound Monitor Project: implementing and assessing analytics for a wireless, battery-less smart sensing platform for wound assessment. The chemotherapy project focused on implementing and improving a variational auto-encoder (VAE) model with additional semi-supervised techniques to maximize performance (area under the receiver operating characteristic, AUROC) for predicting chemotherapy response for five types of cancers (bladder, colorectal, breast, sarcoma, and pancreatic). The wound project's goal is to develop a wearable real-time wound monitor. It is a collaboration involving investigators in electrical engineering, bioengineering, computer science, and veterinary medicine at Oregon State University. We have created a wound image segmentation and healing status prediction computer program—based on machine learning techniques (Convolutional encoder-decoder network (SegNet), U-Net, Gaussian Process Regression (GPR), Random Forest regression (RFR), and XGBoost regression)—that can automatically calculate the area of the wound and predict the number of remaining days until full recovery based on a smartphone-acquired digital image of the wound. We measured the performance of segmentation models via the area of overlap (Dice coefficient) between predicted and human-annotated images. We trained regression models for predicting days until full healing using randomized search and grid search based on R^2 coefficient of determination scoring. We obtained unbiased accuracy estimates using an independent test wound images set and compared prediction performance among the regression models.
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Rights Statement
Publisher
Peer Reviewed
Language
Embargo reason
  • Pending Publication
Embargo date range
  • 2021-09-19 to 2022-10-20

Relationships

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details WeiQi2021.pdf 2021-09-19 Public Download