DigiTIL, a Computational Histomorphometric Predictor of Disease Recurrence and Overall Survival for p16-Positive Oropharyngeal Squamous Cell Carcinoma
Abstract
Objective and Rationale: Oropharyngeal squamous cell carcinoma (OPSCC), one type of head & neck (H&N) cancer, has experienced an increase in incidence in the past decades. Human Papilloma Virus (HPV) is the most important cause of OPSCC, representing approximately 80% of the cases, with a prevalence of 16,000 cases per year. The OPSCC tumors that are associated to HPV are usually referred to as p16 positive. In general, these tumors have favorable outcomes, meaning that patients have a good chance of treatment success. Unfortunately, up to 36% of patients develop disease of recurrence or die from the disease despite the application of aggressive treatments. Conversely, a number of patients with favorable outcomes are subjected to aggressive therapy that likely provides no added benefit. Consequently there is a clear and unmet need for biomarkers to help distinguish between p16 positive OPSCC patients at a high risk, who could benefit from therapy intensification, versus low risk patients who could benefit from therapy de-intensification, thereby reducing the adverse effects of some aggressive treatments. There are some clinically accepted risk predictors for p16 positive OPSCC, which are related to tobacco consumption and the size and location of the tumor, but these are not very accurate. Furthermore, OPSCC affects different populations in different ways. For instance, African Americans (AA) with OPSCC have worse survival rates compared to Caucasian Americans (CA). Similarly, Veterans (V) constitute a population at high risk of death and complications during application of therapy, given their high rates of tobacco consumption (40%-100%) and other health conditions. Understanding how the disease affects such populations could help in the development of more specific and accurate prognostic and predictive risk models. Artificial intelligence (AI) and computer vision techniques offer powerful approaches for rigorously interrogating and analyzing digitized tissue slides from OPSCC patients to obtain objective measurements of shape, arrangement and texture of cells and discover cellular patterns and configurations that are not evident to the human eye. The AI-identified cellular patterns, in turn, could provide important hallmarks with regard to disease outcome and treatment response. In a preliminary study involving N=354 p16 positive OPSCC patients, the principal investigator (PI) showed that AI-derived image features obtained from the interaction between cancerous and immune cells on tissue pathology images helped identify those patients who were at a high risk of disease recurrence. In the present project, the PI seeks to build on these promising initial results to further develop and validate a computational tool for risk stratification of p16 positive OPSCC patients using clinical trial datasets. The PI also seeks to investigate possible differences in the disease-specific patterns on tissue images between different populations (e.g., between African/Caucasian Americans and Veterans/Non-Veterans) in turn allowing for creation of more accurate population specific risk models. The PI’s career goal consists of pursuing an academic career in H&N cancer research, leading to a full-time faculty position with a focus on computational imaging for cancer diagnosis and prognosis. He also seeks to develop a translational research program with the goal of developing digital biomarkers that can be deployed within randomized cancer clinical trials for helping in patient selection. In the short term, the PI is interested in working closely with different cooperative groups to design digital biomarker tools to help identify OPSCC patients at high risk, who could need therapy intensification, and low risk, who could benefit from treatment de-escalation. Additionally, he is very passionate about using these tools to identify potential morphologic differences in the H&N cancer phenotype between different populations (AA vs CA and
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 05, 2021
- Source ID
- W81XWH2110160
Entities
People
- Germán Corredor
Organizations
- Case Western Reserve University
- United States Army