Performance changes due to differences among annotating radiologists for training data in computerized lesion detection.

Purpose: The quality and bias of annotations by annotators (e.g., radiologists) affect the performance changes in computer-aided detection (CAD) software using machine learning.