Application of Bayesian Latent Class Model in Determining the Diagnostic Value of Brain SPECT and MRI for Detecting Posttraumatic Olfactory in the Absence of Golden Standard

The sense of smell gives unexplainable quality to human life. The impairment In this sense will create lot of problems. MRI and SPECT are two way of olfactory evaluation that none of the both is not Gold standard. Bayesian latent class model is the correct way to determine the diagnostic value of these tests.
MRI and SPECT tests performed on 63 patients eligible for the study that went from the beginning of July 2011 to the end of September 2012 to hospital Shahid Rahnemoun. The results ,as maximum likelihood function with prior distribution combines, using Markov chain Monte Carlo in winbugs 1.4.3 software. the median of the posterior distribution presented as the parameter estimates. Both dependent and independent conditional models was compared using criterion DIC.
The sensitivity and specificity of MRI to detect abnormal olfactory ,determined in model conditional depends , 58% and 89% respectively and 73% and 84% for SPECT. positive and negative predictive values were calculated for both the test.Convergence chains and goodness of fit of model using time-series charts and Brook–Gelman–Rubin and Bayesian p-value was confirmed. considering the lower criterion DIC for conditional dependence, this model was determined the best fit to the data.
In Latent Class Model, different results are obtained from the when this model is not used. It is better that both conditional dependent and independent model fitted to the data, and finally were compared.