Abstract
Radiologists often spend hours interpreting medical images. We are interested in the role that visual adaptation may play in the perception of structure in radiological scans. To focus our study, we concentrated on breast imaging, where a judgment of breast density is a standard component of the report from a screening mammogram. Images were taken from random sections within breast regions of normal mammograms and displayed on a CRT. Adaptation was measured with a standard asymmetric matching task. In one set of experiments, we asked whether the “unnatural” statistics of radiological images induce changes in the “natural” adaptation state of the observer. For example, the scans have steeper power spectra, and adapting to them produced shifts in the perceived spectrum of filtered noise consistent with adaptation to blur. In a second set, we asked whether adaptation could selectively alter the appearance of different scans. For example, exposure to an image of tissue classified as “dense” could bias an ambiguous target to appear more “fatty” or vice versa. Our results show that observers can rapidly adapt to the image structure in mammograms, and this could potentially be an important factor in the perception and learning of radiological images.