Abstract
Purpose :
Viewing faces is a critical aspect of everyday life. Low vision patients have problems in understanding facial expressions, hampering social communication in all aspects of life, making it important to know how low vision deficits affect the processing of visual information facial expressions. Whole-head brain imaging through electroencephalography (EEG) provides an advanced non-invasive method of determining the visual information processing involved in the interpretation of facial expressions.
Methods :
Stimuli consisted of 2 full-colored face panels of each of 18 persons, each panel displaying either a neutral or happy expression. The subject pressed either a left or right key to indicate which one was the happy face of the pair. The stimuli were then modified to simulate three kinds of visual degradation in low vision patients. A) The contrast of the original stimulus was reduced by a factor of 10. B) Sampling inefficiency was simulated by overlaying the original stimulus with random arrays of missing samples in the form of local gray pixels. C) Patchy field loss was simulated by overlaying the images with randomly-positioned dark patches. EEG was measured with a high-density EGI electrode net at a 500 Hz sampling rate. The resultant EEG signals were pre-processed to remove blinks and other artifacts and analyzed by Principal Components Analysis (PCA) to reveal fixation-related spatiotemporal visual cortical potentials.
Results :
PCA analysis revealed that at least 6 spatiotemporal response components with occipital lobe sources were identifiable across the 4 test conditions, with a wide range of peak latencies (~100 to ~900 ms). The different forms of low vision degradation affected the components of face processing to different extents. While the weight pattern for patchy field loss was similar to that for the full faces, contrast reduction roughly inverted the component weights, and sampling inefficiency produced a weight pattern that was different from the others.
Conclusions :
The simulated masking manipulations were evidently implicating substantially different neural processing pathways from each other. These results show that spatiotemporal PCA can identify a range of visual cortical response components of the visual processing of facial images, in addition to traditional P1, N1, P2 and P3 ERP components.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.