Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


Leo_dissertation_final.pdf (30.35 MB)

ETD Abstract Container

Abstract Header

Prognostic and Predictive Computational Pathology-Based Companion Diagnostics for Genitourinary Cancers

Leo, Patrick J
http://orcid.org/0000-0002-2248-0610
http://rave.ohiolink.edu/etdc/view?acc_num=case1630601324808315

Abstract Details

2022, Doctor of Philosophy, Case Western Reserve University, Biomedical Engineering.
Correctly estimating cancer aggressiveness enables balancing of treatment benefits and risks. Many patients with prostate and bladder cancer will not die from the disease, but due to a shortage of accurate prognostic tools, will be subjected to extensive testing and treatment. This overtreatment stems from an inability to identify which patients' lives will be extended by aggressive interventions and which could safely forgo treatment. To assess cancer aggressiveness, pathologists examine microscopic tumor morphology on slides stained with hematoxylin and eosin. Intra- and inter-reviewer variation and the limits of human perception hamper the reproducibility and prognostic power of this analysis. Computer vision and machine learning approaches avoid these limitations through objective analysis of pathology. In this dissertation, we developed several novel morphological biomarkers for cancer recurrence prognosis and treatment response prediction. One set of these biomarkers characterized nuclear and glandular morphology in terms of nuclear and glandular shape, architecture, and orientation entropy. A second set of features quantified cellular appearance in intratumoral stroma. Lastly, a specific cancer phenotype, cribriform, was automatically identified and its extent used as a biomarker. These biomarkers were validated for prostate cancer biochemical recurrence prognosis in n=675 patients, for progression from active surveillance in n=191 patients, for treatment response prediction in n=160 patients, and for bladder cancer recurrence prognosis in n=151 patients. Methods to increase the robustness of these models enabled validation on datasets drawn from several institutions with diverse tissue appearances. Population-specific tuning of the morphological signatures constructed from these biomarkers further improved their performance. These biomarkers outperformed current methods, adding value even when controlling for clinical covariates. In a head-to-head comparison, a set of these biomarkers outperformed the Decipher genomic test in prostate cancer. In particular, the developed models were consistently most prognostic in low-grade and clinically low-risk patients. Finally, quantitative morphological analysis was able to predict response to additional treatment following radical prostatectomy. This cannot be accomplished by any currently deployed tool.
Anant Madabhushi (Advisor)
Satish Viswananth (Committee Chair)
Michael Kattan (Committee Member)
Robin Elliott (Committee Member)
Sanjay Gupta (Committee Member)
139 p.
Biomedical Engineering
Prostate cancer; machine learning; digital pathology; precision medicine; bladder cancer; prognostic; predictive; companion diagnostic

Recommended Citations

Citations

  • Leo, P. J. (2022). Prognostic and Predictive Computational Pathology-Based Companion Diagnostics for Genitourinary Cancers [Doctoral dissertation, Case Western Reserve University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=case1630601324808315

    APA Style (7th edition)

  • Leo, Patrick. Prognostic and Predictive Computational Pathology-Based Companion Diagnostics for Genitourinary Cancers. 2022. Case Western Reserve University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=case1630601324808315.

    MLA Style (8th edition)

  • Leo, Patrick. "Prognostic and Predictive Computational Pathology-Based Companion Diagnostics for Genitourinary Cancers." Doctoral dissertation, Case Western Reserve University, 2022. http://rave.ohiolink.edu/etdc/view?acc_num=case1630601324808315

    Chicago Manual of Style (17th edition)

case1630601324808315
4
© 2021, all rights reserved.
This open access ETD is published by Case Western Reserve University School of Graduate Studies and OhioLINK.