Our Bayesian approach to incorporate prior knowledge can enhance efficiency in searching of such associations from data. Conclusion: We demonstrated a probabilistic ensemble method to detect robust associations between RP covariates and its potential to improve RP prediction accuracy. The ensembles obtained by incorporating the prior knowledge improved classification performance for the ensemble size 5∼50. The ensemble BN model achieved the maximum sensitivity/specificity of 81%/84% and outperformed univariate dosimetric predictors as shown by larger AUC values (0.78∼0.81) compared with MLD (0.61), V20 (0.65) and V30 (0.70). Results: The LASSO selected the following 7 RP covariates: (1) pre-RT concentration level of α2M, (2) α2M level mid- RT/pre-RT, (3) pre-RT IL6 level, (4) IL6 level mid-RT/pre-RT, (5) ACE mid-RT/pre-RT, (6) PTV volume, and (7) mean lung dose (MLD). A graph ensemble was formed by averaging the most probable graphs weighted by their posterior, creating a Bayesian Network (BN)-based RP risk classifier. Posterior probability distribution of interaction graphs between the selected variables was estimated from the data under the literature-based prior knowledge to weight more heavily the graphs that contain the expected associations.
The number of biological and dosimetric covariates was reduced by a variable selection scheme implemented by L1-regularized logistic regression (LASSO). Dose-volumetric parameters were also more » included as covariates. From each sample the concentration of the following five candidate biomarkers were taken as covariates: alpha-2-macroglobulin (α2M), angiotensin converting enzyme (ACE), transforming growth factor β (TGF-β), interleukin-6 (IL-6), and osteopontin (OPN). Blood serum was collected from every patient before (pre-RT) and during RT (mid-RT). 16 RP events was observed (CTCAE grade ≥2). Methods: We recruited 59 NSCLC patients who received curative radiotherapy with minimum 6 month follow-up.
Purpose: We propose a prior knowledge-based approach to construct an interaction graph of biological and dosimetric radiation pneumontis (RP) covariates for the purpose of developing a RP risk classifier. No approximations are required in the Bayesian approach and the resulting distributions can be input to a decision analysis to obtain expected utility for the decision to launch. Uncertainties, which are represented by probability distributions in the Bayesian approach, are propagated through the model to obtain a probability distribution for O-ring failure, and subsequently for shuttle failure as a result of O-ring failure. Markov more » chain Monte Carlo (MCMC) sampling will be used to sample from the joint posterior distribution of the model parameters, and to sample from the posterior predictive distributions at the estimated launch temperature, a temperature that had not been observed in prior launches of the space shuttle. In this paper, we will re-evaluate the analyses of Dalal et al. Because their analysis was frequentist in nature, probability distributions representing epistemic uncertainty in the input parameters were not available, and the authors had to resort to an approximate approach based on bootstrap confidence intervals. propagated parameter uncertainties through the fitted logistic regression model in order to estimate the probability of shuttle failure due to O-ring failure at the estimated launch temperature of ~30o F. In the second portion of their paper, Dalal et al.
performed a frequentist analysis of the O-ring data, and found that a logistic regression model provided a relatively good fit to the past data. The purpose of their analysis was to show how statistical analysis could be used to provide information to decisionmakers prior to the launch, information that could have been expected to lead to a decision to abort the launch due to the low temperatures (~30o F.) present at the launch pad on the morning of the scheduled launch. Dalal et al performed a statistical analysis of field and nozzle O-ring data collected prior to the ill-fated launch of the Challenger in January 1986.
0 kommentar(er)
