Abstract
Adaptation to radiological images produces strong aftereffects in the perceived texture of images. For example, prior exposure to mammograms classified as dense or fatty tissue causes intermediate images to appear more fatty or dense, respectively. Previously (Kompaniez et al. VSS 2013) we found that these texture aftereffects can also impact performance on the types of visual detection tasks confronting radiologists: finding a simulated lesion within the image. Observers were faster finding targets (Gaussian spots) when they were first adapted to the type of image (dense or fatty) on which they were searching. In the present study we tested for analogous effects on search when adapting instead to the characteristic power spectra of mammograms. These spectra are steeper (slopes ~-3) than that of natural stimuli (~-2), and we showed previously that adaptation to them produces corresponding aftereffects in perceived blur. For the search task, observers first adapted for 5 min to a series of image sections taken from scans of normal tissue, shown either with the original spectrum or filtered to a "natural" spectrum with a slope of -2 (with rms contrast equated). They then searched for a target within either the unfiltered or sharpened images. Targets were again simulated lesions corresponding to bright Gaussian spots (sd = .18 deg), superimposed by adding the luminance to the background. Unlike the effects for texture, in this case reaction times were unaffected by the adaptation. Moreover, mean search times remained similar on the original or filtered backgrounds, even though the latter simulate theoretically complete adaptation to the power spectra. These results suggest that the search improvements with texture adaptation are not a generic consequence of prior exposure to the images, and may reflect greater selectivity of the adaptation for the phase vs. amplitude differences distinguishing the backgrounds from each other or the targets.
Meeting abstract presented at VSS 2014