- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Biomarkers in screening and management of at-risk oral...
Open Collections
UBC Theses and Dissertations
UBC Theses and Dissertations
Biomarkers in screening and management of at-risk oral lesions – a quantitative cytology focus Jeon, James Sanghyun
Abstract
Early detection and intervention of oral premalignant lesions (OPLs) are at the forefront of clinical endeavors for oral cancer control. Quantitative cytology (QC) and loss of heterozygosity (LOH) have shown to be promising prognostic biomarkers for identifying high-risk OPLs. The objectives of this thesis are to optimize the screening utility of the QC algorithm by assessing nuclear morphology and explore if QC or LOH may serve as predictive biomarkers in treatment management of OPLs. In this thesis, we first used 169 brushing samples from 156 patients to examine the comparability between the MoticEasyScan and ClearCyte and 105 samples from 100 patients to construct a morphometric-based QC-model for screening. Random Forest Models and Receiver Operating Characteristic curves were used to determine the cut-offs to identify abnormalities at cellular and case levels. To evaluate LOH and QC as biomarkers for predicting topical treatment response, a set of 28 OPLs from 28 patients were included for analysis. The Fisher’s exact test was used to assess the predictive value of LOH and QC. The results from the MoticEasyScan system are comparable to those from the ClearCyte for QC analysis. We have identified problems in the previous system with staining batch variations and normalization and both have significant impacts on photometric features used for previous software development. To overcome these problems, we developed a new QC model using morphometric only features. To develop a screening test, using 94 morphometric features with 65.7% of abnormal votes for abnormal cells and 2.7% abnormal cells for abnormal cases, the new model (RF94) yielded a 100% sensitivity in the Test set. To triage OPLs for treatment, the model (mRF10), using 10 morphometric features with 67.3% of abnormal votes for abnormal cells and 6.7% abnormal cells for abnormal cases, yielded in 87.5% specificity in the Test set. Using mRF10 model, no significant differences in proportions of the persistent and resolved groups in QC or LOH were identified at the pre-treatment timepoints. LOH may not be sensitive predictive markers for topical treatment of OPLs. In conclusion, QC models using morphometric-based features can potentially be a useful screening test.
Item Metadata
Title |
Biomarkers in screening and management of at-risk oral lesions – a quantitative cytology focus
|
Creator | |
Supervisor | |
Publisher |
University of British Columbia
|
Date Issued |
2022
|
Description |
Early detection and intervention of oral premalignant lesions (OPLs) are at the forefront of clinical endeavors for oral cancer control. Quantitative cytology (QC) and loss of heterozygosity (LOH) have shown to be promising prognostic biomarkers for identifying high-risk OPLs. The objectives of this thesis are to optimize the screening utility of the QC algorithm by assessing nuclear morphology and explore if QC or LOH may serve as predictive biomarkers in treatment management of OPLs.
In this thesis, we first used 169 brushing samples from 156 patients to examine the comparability between the MoticEasyScan and ClearCyte and 105 samples from 100 patients to construct a morphometric-based QC-model for screening. Random Forest Models and Receiver Operating Characteristic curves were used to determine the cut-offs to identify abnormalities at cellular and case levels. To evaluate LOH and QC as biomarkers for predicting topical treatment response, a set of 28 OPLs from 28 patients were included for analysis. The Fisher’s exact test was used to assess the predictive value of LOH and QC.
The results from the MoticEasyScan system are comparable to those from the ClearCyte for QC analysis. We have identified problems in the previous system with staining batch variations and normalization and both have significant impacts on photometric features used for previous software development. To overcome these problems, we developed a new QC model using morphometric only features. To develop a screening test, using 94 morphometric features with 65.7% of abnormal votes for abnormal cells and 2.7% abnormal cells for abnormal cases, the new model (RF94) yielded a 100% sensitivity in the Test set. To triage OPLs for treatment, the model (mRF10), using 10 morphometric features with 67.3% of abnormal votes for abnormal cells and 6.7% abnormal cells for abnormal cases, yielded in 87.5% specificity in the Test set. Using mRF10 model, no significant differences in proportions of the persistent and resolved groups in QC or LOH were identified at the pre-treatment timepoints. LOH may not be sensitive predictive markers for topical treatment of OPLs.
In conclusion, QC models using morphometric-based features can potentially be a useful screening test.
|
Genre | |
Type | |
Language |
eng
|
Date Available |
2022-06-21
|
Provider |
Vancouver : University of British Columbia Library
|
Rights |
Attribution-NonCommercial-NoDerivatives 4.0 International
|
DOI |
10.14288/1.0415034
|
URI | |
Degree | |
Program | |
Affiliation | |
Degree Grantor |
University of British Columbia
|
Graduation Date |
2022-11
|
Campus | |
Scholarly Level |
Graduate
|
Rights URI | |
Aggregated Source Repository |
DSpace
|
Item Media
Item Citations and Data
Rights
Attribution-NonCommercial-NoDerivatives 4.0 International