Brain energy and discomfort.
(1 CET POINT)
The human visual system evolved to process images from nature. Despite their obvious variation, images from nature have a consistent spatial structure. The fine detail in an image usually has a low contrast relative to the content at larger scales. The Fourier spectrum decreases in amplitude with increasing spatial frequency. In fact, the decrease is approximately proportional to the reciprocal of spatial frequency. As a result, in most natural images, a plot of the Fourier amplitude against spatial frequency on log-log coordinates has a slope close to -1 (see Figure 1). (1-4)
Given that the visual system has adapted to process natural images, one might expect that images with the spatial structure shown in Figure 1 would be easy for the visual system to process; this expectation is borne out in several ways.
Natural images are comfortable
Juricevic et al asked observers to rate the discomfort from meaningless images composed of filtered noise or randomly disposed randomly sized rectangles. (6) For both categories of image, the discomfort was minimal with a 1 / f Fourier amplitude spectrum--that is to say, when the slope was -1 on log-log coordinates. The central pattern in Figure 2 has a slope of -1. The discomfort increased when the slope was greater or less than -1, as in the flanking patterns of Figure 2. So images with a spatial structure similar to that occurring in nature were rated as more comfortable to look at, even though all the rated images were meaningless. Evidently there is something structural about discomfort from images, quite independently of what the images represent.
It is not simply the slope of the Fourier amplitude spectrum that is critical in determining discomfort, but its shape. This was discovered when people were asked to rate modern art. As might be expected, they rated art with a 1 / f spectrum as comfortable to look at. The uncomfortable art had a spectrum that departed from 1 / f, but it did so not in terms of the slope but because of a relative excess of contrast energy at mid-range spatial frequencies. (7) The human visual system is generally most sensitive to mid-range spatial frequencies, those within an octave of three cycles per degree (CPD), (8) so the uncomfortable images had an excess of contrast energy where the visual system is most sensitive. Using artificial images made by filtering random noise, it was shown that departures from 1 / f are responsible for discomfort if the departure occurs at spatial frequencies close to three cycles/ degree; (7) this applies even when the apparent contrast is controlled. (9) Generating new images by exchanging the phase and amplitude of comfortable and uncomfortable images has shown that the discomfort is determined by the Fourier amplitude rather than the phase information. It is the amplitude that determines discomfort, even though the phase information is generally more useful in identifying objects. (7)
The Fourier amplitude spectrum is two-dimensional it reflects the periodicity of the images at all orientations (vertical, horizontal and all orientations in between). The studies described above measured the Fourier amplitude spectrum by averaging over all orientations, as is customary. Averaging over orientations loses the distinction between periodicity in one orientation and that in another. Checkerboards (which have contrast energy in several orientations) are less uncomfortable than stripes in which the energy varies only in one orientation. (10) Penacchio and Wilkins, therefore, measured the Fourier amplitude in two dimensions. (11) Instead of averaging over all orientations and fitting a straight line on log-log coordinates as had previously been done, they fitted a cone with slope of -1 to the two-dimensional log amplitude spectrum. The residual error in the fit provided an index of how close to 'natural' the Fourier spectrum of the image was, and this turned out to predict ratings of discomfort from the images. Seven sets of images were analysed (765 images in all), and each set had a very different character: photographs of everyday scenes; of buildings; of animals; randomly generated polka dots; and non-representational art.
Despite the large range of images, the index explained 17% of the variance in judgments of discomfort. The prediction was improved when the residuals were weighted to take account of the greater sensitivity to mid-range spatial frequencies, as in a published estimate of the contrast sensitivity function. (12) From these two principles obtained entirely from the literature (without fitting any parameters) they were able to explain an average of 27% of the variance in judgments of discomfort.
[FIGURE 1 OMITTED]
In summary, two related factors have been found to predict judgments of discomfort from images: departure from the statistics of natural images; and excess energy at the spatial frequencies to which the human visual system is generally most sensitive. So the more unnatural an image is and the greater its contrast energy at mid spatial frequencies, the more uncomfortable the image.
Natural images can be processed efficiently
As mentioned earlier, images from nature have a 1 /f structure. The human contrast sensitivity function is optimised for encoding images with this structure. (13,14) The receptive fields of neurons in the primary visual cortex are such that images with 1 / f structure produce a 'sparse' cortical response. (13,15) The defining characteristic of this 'sparse' response is that at any one time few neurons are active while many are inactive, thereby reducing metabolic demand. So the design of the receptive fields may minimise metabolic demand.
Hibbard and O'Hare have built a simple computational model of visual area VI. The model shows that uncomfortable stimuli such as striped patterns, which are rare in nature and do not conform to a 1 / f structure, give rise to an excess of 'neural activity' and a non-sparse distribution of 'neural' firing."' Their finding has been confirmed and extended using a more elaborate model that includes the excitatory and inhibitory connections between neurons. According to this model, images rated as uncomfortable to look at give i rise to a less sparse response. (17)
[FIGURE 2 OMITTED]
The computational models of the visual cortex mentioned above suggest that uncomfortable and unnatural images involve a less sparse coding and a greater neural response overall. As will now be shown, this prediction is supported by physiological evidence.
Uncomfortable images, oxygenation and homeostasis
When a visual stimulus is observed, there is a change in the oxygenation of the blood reaching the visual cortex--the cortical haemodynamic response. This response to visual stimuli reflects the activity of large numbers of neurons and their local collective demand for oxygenated blood.
The relationship between the amplitude of the haemodynamic response and the size of the neuronal response is complex and indirect. It is affected by many factors such as blood flow and glial cell activity, but generally it broadly reflects local field potentials. (18) The response can be measured using functional magnetic resonance imaging (fMRI) or near infrared spectroscopy (NIRS) in which infrared light penetrates the scalp and measures the colour of the surface of the brain. As we will now see, both techniques show that the oxygenation is greater when the visual stimulus is uncomfortable.
Huang et al measured the fMRI Blood Oxygen Level Dependent (BOLD) response to achromatic gratings with a range of spatial frequencies and showed that those with mid spatial frequency (which are uncomfortable) gave the largest response. (19) It is generally the case that individuals who are susceptible to discomfort show a larger BOLD response than those who are not. Patients with migraine who report relatively high levels of discomfort from patterns that give a BOLD response j with relatively high amplitude. (20,21) The number of I symptoms of discomfort reported by patients who experience migraine with aura correlates with the amplitude of the BOLD response to visual stimulation. (22)
The relationship between discomfort and the size of the haemodynamic response can occur independently of the diagnosis of migraine. Thus, Alvarez-Linera Prado et al compared 20 photophobic patients with 20 controls who viewed a light source at various intensities. (23) There was a direct relationship between stimulus intensity and the size of the BOLD response, and the response was higher in the photophobic patients, particularly at low and medium light intensities. Bargary et al compared normal participants with high and low discomfort glare thresholds while they identified the orientation of a Landolt C surrounded by peripheral sources of glare. (24) The group that was sensitive to discomfort glare had an increased BOLD response localised at three discrete bilateral cortical locations: in the cunei, the lingual gyri and in the superior parietal lobules.
There is a relationship between discomfort and the magnitude of the haemodynamic response in the visual cortexin terms of the stimuli that evoke discomfort, which generally induce a large response, and in terms of the individuals who are susceptible to discomfort, in which the response is larger than in others.
It is possible that the discomfort is homeostatic. As with any other pain, it encourages withdrawal and may act to reduce the use of energy by the brain. The brain constitutes 2% of body weight but consumes 20% of the body's energy. Only a small fraction, perhaps 1%, of the cerebral cortex can be supplied with energy and be active at any given time, (25,26) so conservation of metabolic energy is an important requirement.
Colour and discomfort
So far we have considered mainly patterns that vary in luminance. Similar considerations seem to apply to patterns that differ only or mainly in colour. Haigh et al measured the discomfort from gratings with bars that alternated between two colours. (27) They showed that discomfort from these patterns was predicted from the difference in colour (the separation of the chromaticities in the CIE 1976 UCS diagram): the larger the difference, the greater the discomfort and the larger the haemodynamic response. This was the case in five studies and for a large gamut of colours, some with different luminance. Juricevic et al also showed that discomfort was greater for images with a large colour difference. (6) They used images comprising random dots or randomly disposed rectangles and measured the colour difference in terms of the L-M and S-LM chromatic plane. Large colour contrasts are rare in the natural world, (6,28) so, in both the studies, discomfort was associated with images that are rare in nature. It is possible that tinted lenses can reduce discomfort partly because, on average, they reduce the colour contrasts in the visual scene.
Implications for design
Comparison of rural and urban images using the algorithm of Penacchio and Wilkins clearly shows that the image structure of urban scenes is further from 1 /f. In the modern urban environment images with the spatial characteristics of uncomfortable images are commonplace. Everywhere you look there are spatially repetitive patterns, sometimes as a result of design, sometimes because it is cheaper to construct objects from similar components. In a recent study we took photographs of buildings in various UK cities simply by pointing a camera across the street; these photographs were analysed and rated. As might be expected, we found we could predict ratings of discomfort on the basis of the spatial structure of the photographs--and those photographs that were uncomfortable gave rise to a larger cortical haemodynamic response. Analysis of images of apartment buildings built during the last century shows an increasing departure from 1 / f structure with each decade. Our visual environment appears to be getting progressively more uncomfortable.
Discomfort matters. It affects health. The visual stimuli that provoke discomfort are also those that provoke migraine and photosensitive epilepsy. (10,29)
The literature is beginning to form a coherent picture. We are beginning to understand why some things are uncomfortable to look at. Uncomfortable images are processed inefficiently, increasing neural activity and metabolic load in the visual cortex. These uncomfortable images are commonplace in the urban environment.
Under the enhanced CET rules of the GOC, MCQs for this exam appear online at www.optometry.co.uk. Please complete online by midnight on 11 November 2016. You will be unable to submit exams after this date.
CET points will be uploaded to the GOC within 10 working days. You will then need to log into your CET portfolio by clicking on 'MyGOC' on the GOC website (www.optical.org) to confirm your points.
Visit www.optometry.co.uk, and click on the 'Related CET article' title to view the article and accompanying 'references' in full.
Course code: C-52578 Deadline: 11 November 2016
* Be able to advise susceptible patients about how to reduce exposure to uncomfortable visual stimuli (Group 1.2.4)
* Be able to recognise patients that are susceptible to abnormal neurological responses to specific visual stimuli (Group 6.1.14)
Professor Arnold J Wilkins DPhil, FBPsS, CPsychol, HonFCOptom
Professor Arnold Wilkins obtained a doctorate from Sussex University for work on human memory and spent two years as a post-doctoral researcher at the Montreal Neurological Institute where he became interested in photosensitive epilepsy. He then spent 22 years at the MRC Applied Psychology Unit in Cambridge where he showed that fluorescent lighting can cause headaches and he developed a system for precision ophthalmic tinting. He moved to a chair at Essex University in 1997 and continues to work on photophobia research.
|Printer friendly Cite/link Email Feedback|
|Author:||Wilkins, Arnold J.|
|Date:||Oct 1, 2016|
|Previous Article:||Managing oculoplastic concerns in patients with glaucoma.|
|Next Article:||Non-infectious uveitis--current therapies and drugs in the pipeline.|