Despite significant advances in disease prevention and screening, locally-advanced cervical cancer (LACC) continues to be an important world-wide public health problem. Responses to currently available treatments are highly variable among patients with similar constellations of accepted prognostic features, implying that chemo-radio sensitivity, patterns of local tumor proliferation, and the propensity of metastatic spread are highly dependent on each patient's potentially unique cervix cancer habitat. Radiomic analysis of pre-, in-, and post- treatment medical images is an exciting new class of prognostic cancer biomarkers for characterizing tumor habit and has shown promise in more accurately separating favorable- from unfavorable- prognosis for patients for several tumor sites. No such radiomic-based prognostic model has been developed, for LACC that supports meaningful personalization of local and systemic treatment or biologically treatment adaptation based on response of radiomic biomarker profiles to the initial fractions of radiotherapy. To support future personalized chemo-radiation therapy, we propose to exploit a large, prospectively collected database of ~400 LACC patients consisting of archived PET, MRI, and CT images, standard biomarkers, and clinical outcomes to develop an innovative and accurate radiomics-based prognostic model for predicting LACC treatment outcomes. Previous efforts to utilize high-dimensional radiomic feature spaces to predict cancer treatment outcomes have been compromised by small patient numbers relative to the feature space dimensionality; radiomic feature redundancy, feature heterogeneity and uncertainty; unbalanced outcome class patient cohorts; and intrinsic differences in prognostic information content for features extracted from different modality images. We propose a principled and systematic strategy to address these challenges: a novel combination of belief function theory (BFT) and more standard machine learning techniques. To achieve these goals, we propose the following specific aims.
Aim 1 : Develop a novel radiomics-based prognostic model for predicting outcomes of cervical cancer patients treated with definitive chemo-radiotherapy given longitudinal radiomic features from pre-, in-, and post treatment PET, CT, and MR images.
Aim 2 : Refine the developed prognostic model by use of the full 400 patient cases and systematically validate and test its performance.
Aim 3 : Investigate the potential clinical impact of the radiomic-based prognostic model and selected clinical applications. Clinical Impact and future plans: This project will result in a clinically useful family of models for predicting disease control outcomes from current LACC treatments. This clinical decision-making tool can be used to personalize initial treatment prescriptions as well as adaptive treatment prescription when combined with in-treatment radiomic biomarkers. We anticipate that many of the proposed methodological improvements investigated by this proposal will improve tumor response assessment in other disease sites.

Public Health Relevance

This project will result in a clinically useful family of models for predicting disease control outcomes from current locally advanced cervical cancer treatments. This clinical decision-making tool can be used to personalize initial treatment prescriptions as well as adaptive treatment prescription when combined with in- treatment radiomic biomarkers. We anticipate that many of the proposed methodological improvements investigated by this proposal will improve tumor response assessment in other disease sites.

Agency
National Institute of Health (NIH)
Institute
National Cancer Institute (NCI)
Type
Exploratory/Developmental Grants (R21)
Project #
5R21CA223799-02
Application #
9727958
Study Section
Radiation Therapeutics and Biology Study Section (RTB)
Program Officer
Zhang, Yantian
Project Start
2018-06-19
Project End
2020-05-31
Budget Start
2019-06-01
Budget End
2020-05-31
Support Year
2
Fiscal Year
2019
Total Cost
Indirect Cost
Name
Washington University
Department
Radiation-Diagnostic/Oncology
Type
Schools of Medicine
DUNS #
068552207
City
Saint Louis
State
MO
Country
United States
Zip Code
63130