The Library
Deep learning based frameworks for patient selection
Tools
Awan, Ruqayya (2022) Deep learning based frameworks for patient selection. PhD thesis, University of Warwick.
PDF
WRAP_Theses_Awan_2022.pdf - Submitted Version Embargoed item. Restricted access to Repository staff only until 16 June 2024. Contact author directly, specifying your specific needs. - Requires a PDF viewer. Download (17Mb) |
Official URL: http://webcat.warwick.ac.uk/record=b3821807
Abstract
Recently, deep learning (DL) has become a spearhead for solving many problems in the computer vision domain, with computational pathology (CP) has no exception. In the CP domain, it is widely used for histological assessment of tissue for diagnosis and prognosis of cancer patients. The research community has developed an abundance of DL based CP tools, reporting state-of-the-art results, for many diverse applications. In near future, we can envisage better tools on the way forward to the clinical workflow to assist pathologists in making diagnostic and anti-cancer therapeutic decisions. In this thesis, we develop DL based frameworks defining eligibility criteria for selecting patients of two different types of cancers: bladder and colorectal cancers.
We develop our first framework with the main goal to investigate an automated alternative to risk stratification of urine cytology slides. We utilised digital cell profiles for the identification of patients with low-risk and high-risk of developing bladder cancer. Our experiments demonstrate that the digital risk could be a better predictor of the final histopathology based diagnosis. We then develop our second framework for the assessment of mismatch repair (MMR) status to identify patients with microsatellite instability (MSI), known to respond well to immunotherapy. We perform multi-stain tissue analysis using slides stained for MMR protein, in addition to H&E and cytokeratin stained slides. To the best of our knowledge, it is the first time that MMR status is utilised for MSI prediction. Registration is an important pre-requisite task before this multi-stain slide analysis. To this end, we present two approaches utilising two different features, hand-crafted and data-driven features. We adopted a multi-scale and multi-stage strategy, important for improving the quality of registration. These methods are able to align the images with low registration error as compared to other hand-crafted based approaches.
Item Type: | Thesis (PhD) | ||||
---|---|---|---|---|---|
Subjects: | Q Science > QA Mathematics > QA76 Electronic computers. Computer science. Computer software R Medicine > RB Pathology R Medicine > RC Internal medicine |
||||
Library of Congress Subject Headings (LCSH): | Deep learning (Machine learning), Computational biology, Pathology -- Data processing, Medical informatics, Histology, Pathological, Imaging systems in medicine, Diagnostic imaging | ||||
Official Date: | June 2022 | ||||
Dates: |
|
||||
Institution: | University of Warwick | ||||
Theses Department: | Department of Computer Science | ||||
Thesis Type: | PhD | ||||
Publication Status: | Unpublished | ||||
Supervisor(s)/Advisor: | Rajpoot, Nasir M. (Nasir Mahmood) | ||||
Sponsors: | University of Warwick. Chancellor's International Scholarship | ||||
Format of File: | |||||
Extent: | xxi, 128 leaves : illustrations | ||||
Language: | eng |
Request changes or add full text files to a record
Repository staff actions (login required)
View Item |