If you have ever taken an intro class to sensation and perception, chances are that you have experienced the “waterfall illusion” as an example of motion aftereffects. You can try this yourself: Next time you stumble upon a waterfall, stare at it without moving your eyes for about a minute. When you then look at some stationary rocks nearby, these rocks appear to be moving upwards slightly. The illusory upward movement is your experience of the motion aftereffect. If you don’t have a waterfall handy, you can try this video here.
Generally speaking, such perceptual aftereffects can be understood as a response to a stimulus that is unchanging in a particular attribute (e.g., color, movement, orientation, temperature). By staring at this unchanging attribute for a while, our subsequent perception of that attribute, or our capacity to perceive it, is briefly altered (Mollon, 1974). Adaptation is the process by which the perceptual system modifies its operating properties in response to changes in the environment, resulting in perceptual aftereffects. These have become to be known as the “psychologist’s micro-electrode” (Blakemore, 1973; Frisby, 1980).
In a recent study published in Attention, Perception, & Psychophysics, researchers Kompaniez-Dunigan, Abbey, Boone, and Webster examined the influence of sensory adaptation on visual search. In particular, they were interested in whether the salience of target features, bright spots superimposed onto x-ray mammograms mimicking “lesions”, could be enhanced by prior adaptation to the spatial structure of the images. The authors had previously shown that adaptation to medical images can produce robust and rapid aftereffects in the perceived texture of mammogram images.
Why would one want to adapt to mammograms? Visual adaptation usually implies that looking at a set of features for an extended time decreased sensitivity to these and makes them less perceivable. Following this logic you could make your search for the needle in the haystack easier by adapting to the hay in the hope it vanishes from your perception and leaves only the needle.
Mammograms are one of the many “haystacks” that radiologists have to search through on a daily basis. Since the outcome of their searches has vital consequences, helping them find the needle could eventually save lives. So could adapting to the texture of mammograms help detect abnormalities (simulated masses) within them?
To test this, Kompaniez-Dunigan and colleagues randomly selected sections of images taken from mammograms classified as either dense or fatty. These served as the adapting stimuli. The targets were simulated masses corresponding to bright fuzzy spots, each of which was superimposed randomly onto a subsection of a mammogram, with contrast varied over five levels so that detectability varied from difficult to easy. The picture below shows two sample stimuli:
In the search task, a test image was randomly selected from either the dense or the fatty set and participants (who were not radiologists) had to indicate as quickly as possible whether the target fell on the left or the right side of the monitor screen.
Before searching, observers initially adapted either to a uniform gray field for 30 s or they spent 5 minutes adapting to either fatty or dense images, during which the participants were free to move their eyes and explore the adapting sequence without maintaining fixation. A 4-second re-adapt period to the gray screen or a new random sequence of the adapting images was inserted between search trials.
Kompaniez-Dunigan and colleagues found that prior exposure to dense or fatty images facilitated search for simulated masses (those fuzzy spots) embedded in mammogram images. Participants were on average 190 ms faster and 13% more accurate after adaptation. This enhancement was specific to the type of adapting image (dense versus fatty), implying that the observed enhancement was not due to simple generic learning, but that adaptation effects were selective to either “fatty” or “dense” images, i.e. they “were selective for the specific characteristics that distinguished the dense and fatty images.“ This selectivity is consistent with the selectivity in the appearance of the mammograms after adaptation observed by the authors in their previous study (Kompaniez et al., 2013).
This study therefore shows that adapting to the background characteristics of a search display can increase the signal-to-noise ratio of the target within its context to boost detection performance. So no more need to search the haystack? Do radiologists now just need to watch random mammogram textures to improve their performance? Not exactly; it is a bit more nuanced than that. However, the authors suggest ways in which mammogram reading routines could be altered to make use of adaptation.
For instance, if it turns out that adaptation effects persist across case reading times, the sequence of cases could be ordered such that cases of similar tissue densities are viewed together rather than being intermixed with others of dissimilar density. This could potentially not only increase search efficiency, but might also reduce fatigue. In a previous paper, Webster explored opportunities “to adapt images instead of observers”, in order to “simulate how images should appear under theoretically optimal states of adaptation”. Depending on the ability to predict adaptation, images from new environments could be “pre-adapted” to fit specific observers, eliminating the need for observers to adapt. This sounds like it might be worth exploring further, although what a pre-adapted haystack would look like remains to be seen.