Reader variability: what we can learn from computer-aided detection experiments.

Abstract

Radiologists interpreting images vary in their identification of disease. When 1 radiologist reinterprets the same case differently, it is called intraobserver variability; when 2 radiologists differ with each other on a case, it is called interobserver variability. Computer-aided detection (CAD) systems can increase the detection of disease. When radiologists use these systems, they are aided in the detection of lesions that they might have detected had they reinterpreted the images or that other radiologists might have identified had they interpreted the images without CAD. This article demonstrates how most cases "newly" identified by radiologists working with CAD are actually cases that the radiologists or other radiologists would have identified had they interpreted the images without CAD. Computer-aided detection, therefore, decreases intraobserver variability and interobserver variability.