|
Activity Number:
|
76
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #304179 |
|
Title:
|
A Bayesian Generalized Non-Linear Predictive Model of Treatment Efficacy Using qMRI
|
|
Author(s):
|
Jincao Wu*+ and Timothy D. Johnson
|
|
Companies:
|
University of Michigan and University of Michigan
|
|
Address:
|
1952 Traver Road Apt 202, Ann Arbor, MI, 48105,
|
|
Keywords:
|
qMRI ; Bayesian ; prediction ; MARS
|
|
Abstract:
|
The prognosis for patients with high-grade gliomas is poor with a median survival of one-year post diagnosis. Researcher hypothesize that quantitative MRI (qMRI) can reduce assessment of treatment efficacy from 8-10 weeks post treatment to 3 weeks post therapy initiation, thereby allowing second line treatments to begin earlier. The purpose of this work is to build a predictive model for the treatment efficacy based on qMRI data. We use 1 year survival status as the outcome and propose a Bayesian joint model, iterating between: 1) smoothing the qMRI data using a pairwise-difference prior and deriving summary statistics; 2) fitting a generalized non-linear model in the Bayesian setting, where statistics from stage 1 are covariates. Multivariate Adaptive Regression Splines are used as basis functions in the model. Bayesian model averaging is employed to derive the final predictive model.
|
