Retinal neural cells sample images of the external world, convert light energy into neural signals, and transmit neural signals to the brain. The integrity of the retinal neural cells is crucial to normal spatial vision. Many retinal diseases, such as age related macular degeneration, diabetic retinopathy and retinitis pigmentosa, are known to cause degeneration of retinal neural cells. Former Vision Research Investigator Dr. Lei Liu and Senior Fellow in Vision Science Aries Arditi started a series of psychophysical experiments which aim to reveal the relationship between the integrity of retinal neural cells and spatial vision performance (visual acuity, contrast sensitivity and texture discrimination).

In these studies, a computer graphics technique is used to generate degraded visual stimuli that mimic images sampled by retinas that suffer from neural loss. These stimuli are then used to test both normal subjects and patients with known retinal diseases. The performance of the normal subjects serves as the reference baseline for assessing tolerance to image degradation. It is expected that the patients with retinal diseases will show poorer tolerance to image degradation because the defects of their retinas will compound the defects of the image. By comparing the performance of normal subjects with that of patients, the magnitude of a patient's retinal neural loss may be estimated.

The first experiment of this project measured normal subjects' tolerance to sample noise, luminance noise and positional uncertainty in texture discrimination tasks. Specially designed texture pairs were used in this study. Subjects with normal vision could easily discriminate intact texture pairs. When more pixels of the texture patches were blanked out or more luminance noise was added to the texture patches, the discrimination became more difficult; thus, a threshold tolerance to missing pixels or luminance noise could be measured. Drs. Liu and Arditi found that normal subjects needed only 40% of the pixels in the texture patch to achieve 82% correct texture discrimination. This result indicated that normal subjects could tolerate a large amount of luminance noise. Discrimination performance for normal subjects remained high as long as the contrast polarities of the majority of the pixels remained unaltered. In contrast, they found that if the remaining pixels in a sampled texture patch were allowed to move into vacancies made by their deleted neighbors (positional uncertainty), then normal subjects' texture discrimination performance was greatly impaired. Under this condition, normal subjects required about 80% of pixels to reach 82% correct texture discrimination. These findings appeared in the proceedings of an Optical Society of America topical meeting (Liu, 1999b) and in abstract form (Liu & Arditi, 1998; Liu, 1999a).

The degraded texture discrimination test (DTDT) has been applied to two retinal diseases, AMD and diabetic retinopathy to evaluate its value as a tool to detect and quantify neural loss associated with early retinal diseases. With the support of the Juvenile Diabetes Research Foundation and the Helen Hoffritz Charitable Trust, and in collaboration with Dr. Janis White of the VA hospital in East Orange, NJ, Dr. Liu conducted DTDT on young and old normal controls, diabetic patients without retinopathy, diabetic patients with early retinopathy, and patients with early AMD. These experiments have demonstrated that DTDT can differentiate patients with and without retinal diseases. Specifically, in DTDT, diabetic patients with early retinopathy showed inferior performance to diabetic patients without retinopathy, and patients with early AMD showed inferior performance to age-matched normal controls. The fact that all normal controls and patients had normal visual acuity indicates that DTDT can provide information that cannot be provided by conventional vision test. Therefore, DTDT can be a useful tool in detecting and quantifying early retinal diseases. These findings were presented in several conferences.

A second experiment compared normal subjects' and patients' performances in detecting degraded visual stimuli. In this study, the minimum contrast levels required to detect intact and degraded visual stimuli (the contrast thresholds) were measured. Based on theoretical analyses, Dr. Liu predicted that contrast sensitivity (the reciprocal of contrast detection threshold) should decrease linearly with increasing proportion of stimulus pixel deletion. However, if one could imagine a line graph depicting this relationship, the line would be predicted to have a smaller slope if a fixed proportion of stimulus energy were lost when the textures were imaged on a diseased retina. Dr. Liu measured contrast detection thresholds at several levels of stimulus pixel deletions. Indeed, he observed the predicted linear reduction of contrast sensitivity with increasing proportion of pixel deletion. In addition, when the contrast sensitivity data from the eyes of a patient with advanced retinitis pigmentosa was plotted on a line graph, the slopes of the lines were smaller than the lines depicting the data from normal subjects. The slope differences suggested a 25% and a 35% loss of retinal samples in the 20/20 eye and the 20/30 eye of the patient, respectively. These findings appeared in abstract form (Liu, 1999a).

Further Reading

Liu, L., & White, J. (2003). Using degraded texture discrimination test to detect early age-related macular degeneration. Poster presented at the Annual meeting of the Association for Research in Vision Ophthalmology, Ft. Lauderdale, FL.

Liu, L., & White, J. (2002). Using degraded texture discrimination test to detect early diabetic retinopathy. Optometry and Vision Science (suppl.), 79, 284.

Liu, L., & White, J. (2002). A computational model for discrimination of even and random textures. Fall Vision Meeting, October 24-27, San Francisco, CA (in cooperation with the Optical Society of America).

Liu, L. (2002, May). What makes even textures different from random texture? Paper presented at the annual meeting of the Association for Research in Vision and Ophthalmology, Ft. Lauderdale, FL.

Liu, L. (1999a). Contrast sensitivity and the integrity of retinal photoreceptor array. Investigative Ophthalmology & Visual Science (Suppl.), 40, S36.

Liu, L. (1999b). Texture discrimination is robust to image sampling and noise, but is sensitive to image scrambling. Vision science and its applications, OSA technical digest series (pp. 2-5). Washington, DC: Optical Society of America.

Liu, L., & Arditi, A. (1998). The effect of random pixel loss and multiplicative noise on texture discrimination. Investigative Ophthalmology & Visual Science (Suppl.), 39, S860.



Like us on Facebook   Follow us on Twitter   Watch us on Youtube   Email   Visit our Flickr