Activity Number:
|
246
- Improved Disease Classification Through Extensions of ROC Curve Estimation and Biomarker Characterization
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 9, 2022 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Section on Medical Devices and Diagnostics
|
Abstract #322957
|
|
Title:
|
Diagnostic Latent Class Model for Ordinal Classification
|
Author(s):
|
Xiaoyan Lin* and Yun Yang and Kerrie Nelson
|
Companies:
|
University of South Carolina and University of South Carolina and Boston University School of Public Health
|
Keywords:
|
binormal;
bias;
magnifier;
disease severity;
ordinal;
ROC
|
Abstract:
|
Many disease diagnoses involve subjective judgments by qualified raters. It is important to evaluate individual and overall diagnostic accuracy of these raters. In this paper, we focus on ordinal classification processes. A Probit ordinal hierarchical model is proposed linking rater’s ordinal ratings with rater diagnostic skills (bias and magnifier) and patient latent disease severity. Each patient latent disease severity specifically assumes following a latent class normal mixture distribution. Diagnostic parameters of each rater and latent disease severity and disease binary status of each patient are estimated via an MCMC algorithm. Our model specification provides close form of overall and individual rater’s Receiver Operator Characteristic (ROC) curves and of the area under these ROC curves (AUC), which further facilitates the traditional diagnostic accuracy analysis. Furthermore, the model is extended by incorporating rater covariate effects for the diagnostic skill parameters and patient covariate effects for the latent disease severity. Simulation studies are conducted to evaluate the proposed methods, and the methods are illustrated with a mammography example.
|
Authors who are presenting talks have a * after their name.