NSF/FDA SIR: Quantitative Decision Analysis and Utility Assessments for Medical Imaging Technology

本文以筛查乳房X光摄影为例，研究了基于效用的诊断性能测量方法的发展和分析，以达到技术评估的目的。这种方法与标准分析方法有很大的不同，标准分析方法使用受试者工作特性(ROC)曲线分析来获得曲线下面积(AUC)作为研究终点。实用方法将注意力限制在ROC曲线上最相关的部分。这导致了更有意义的性能度量，也减少了相对于效应大小的外部变量的来源，从而导致了更强大的实验统计设计。

ABSTRACT#1238502AbbeyThis proposal investigates the development and analysis of utility based measures of diagnosticperformance for the purpose of technology evaluation, using screening mammography as an example.This approach represents an important departure from the standard analysis, which uses receiver operatingcharacteristic (ROC) curve analysis to obtain area under the curve (AUC) as the study endpoint. Theutility methods confine attention to the most relevant part of the ROC curve. This leads to amore meaningful measure of performance, and also reduces sources of extraneous variabilityrelative to effect size, thereby leading to a more powerful statistical design of experiments.