Radiographic Techniques and Lesion Detection Performance in Digital Mammography

摘要

Our experimental work uses a Lorad Digital Mammography system, together with an anthropomorphic breast phantom. We have developed techniques to produce a-digital version of masses and calcifications, which will possess an adjustable level of contrast. A 4 Alternate Forced Choice (4-AFC) methodology has been implemented to objectively measure imaging performance, and which has been shown to be markedly superior to subjective methods of assessment of imaging performance. A Monte Carlo code has been developed which enables theoretical calculations to be performed on the expected improvement in observer performance with changing radiographic techniques and phantom thickness. We have performed the following experimental studies: (1) use of an ACR phantom to assess dose & image quality in digital mammography; (2) Observer performance as a function of radiographic technique factors; (3) The importance of size and random noise on lesion detection performance in digital mammography; (4) The effect of lesion location affect lesion detection in digital mammography, which includes both breast thickness and structured (anatomical) background effects. Our results to date indicate that technique factors are of little importance for lesion detection in breast phantoms, but that breast thickness and structured background do affect lesion detection.