Details

Autor(en) / Beteiligte

• Yankeelov, Thomas E
• Hormuth, David A
• Jarret, Angela M
• Lima, Ernesto
• McKenna, Matthew T
• Oden, JT

Titel

Quantitative imaging to enable practical computational modeling of cancer

Ist Teil von

• The Journal of nuclear medicine (1978), 2018-02, Vol.59 (2), p.377

Ort / Verlag

New York: Society of Nuclear Medicine

Erscheinungsjahr

2018

Quelle

EZB Free E-Journals

Beschreibungen/Notizen

• Three fundamental challenges in mechanism-based modeling of cancer are: 1) model construction and selection, 2) estimating model parameters and quantifying observational data, and 3) assessing model prediction and utility. From each challenge, we pose a fundamental question. First, given the enormous number of models covering a huge range of physical and biological events, which models are the "best" for predicting quantities of interest in tumor growth? Second, how can we determine reasonable values for the multitude of model parameters-and their associated errors-that appear in models of tumor growth? Third, how is the uncertainty in the predicted quantities of interest characterized, and how can the model predictions significantly improve patient care? We have established a general framework for addressing each of these challenges. For challenge 1, we develop a large family of multi-scale models that employs a combination of continuum mixture theory and reaction-diffusion equations to describe interacting species, the balance laws of continuum physics, and the principal hallmarks of cancer (1). After initializing and constraining the model family with relevant in vitro and in vivo data (challenge 2), we then overlay our Occam Plausibility Algorithm (OPAL) to systematically select the most appropriate model for predicting future tumor and treatment response (challenge 3). OPAL is a novel and powerful Bayesian approach for bringing together model-specific experimental data for parameter calibration, determination of output sensitivities to parameter variances, calculation of model plausibility for model selection, and development of criteria for designing validation experiments (2). We posit that this approach allows us to develop practical, tumor forecasting methods for predicting the response of individual cancer patients to therapy. We seek to accomplish this goal by integrating advanced, quantitative, multi-modality imaging data (MRI, PET, and microscopy) with multi-scale biophysical models that predict eventual response. In this presentation, we will describe our approach to integrating imaging data with model building and then present early applications of the methodology in various in vitro and in vivo studies involving brain (3,4,5) and breast cancers (6,7). Our overarching hypothesis is that patient-specific quantitative imaging data combined with estimates of therapeutic regimens will enable a multi-scale model to accurately predict responder/non-responder status after a single cycle of therapy on an individual patient basis (8).