A Computational Analysis Pipeline for Imaging Mass Cytometry Data for Cancer Research

Loading...
Thumbnail Image

Authors

Thirumal, Sindhura

Date

Type

thesis

Language

eng

Keyword

imaging mass cytometry , analysis pipeline , phenotyping , deep learning , multiplex imaging , high dimensional data

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

With recent technological advancements in tissue imaging, biologists are now able to obtain a deeper look at tissues at the cell level. Imaging mass cytometry (IMC) is a new advancement in tissue imaging that is quickly gaining wider usage since its recent launch. It improves upon immunofluorescence by quantifying up to 36 discrete signals with the use of metal-tagged antibodies. As a result, the data output of this technology is extensive and complex. The current methods of analysis for these data are provided by the manufacturer and are fairly preliminary, requiring the use of multiple independent software. Studies using this data are typically bounded to simple statistical analyses since the target audience of this technology is biologists with limited computational knowledge. In this thesis, we propose and demonstrate various analysis methodologies with respect to IMC data. First, we develop a software TITAN that is a culmination of the functionalities provided in the current IMC analysis software. TITAN is an improvement on the current methods in efficiency and user-friendliness, as it allows all processing and simple analysis functions to occur within a single environment. Next, we propose an application of machine learning to the analysis pipeline of IMC data, specifically towards cancer research and predicting a clinical outcome. We demonstrate the feasibility of using the proposed framework to identify prognostic features, which is an important foundation for further clinical research. As well, we address the importance of cell phenotyping in relation to IMC analyses and present an automated cell annotation method using deep learning, specifically a convolutional autoencoder with joint classification. We demonstrate the use of these methodologies using a muscle invasive bladder cancer patient cohort. The dataset consists of MIBC patients' transurethral resection of bladder tumour tissues along with their matched cystectomy samples. In addition to demonstrating these methodologies, we also complete a qualitative analysis of this dataset. We compare the TURBT and cystectomy samples qualitatively using a deep autoencoder network and identify similarities between the two that would allow for a pathologist to use TURBT samples as a surrogate for cystectomy samples. Through this work, we demonstrate the potential of using machine learning with IMC data and illustrate its use for cancer research.

Description

Citation

Publisher

License

Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.

Journal

Volume

Issue

PubMed ID

External DOI

ISSN

EISSN