Neurophysiology Study Of Face Perception Under Simulated Prosthetic Vision

Artificial vision studies have validated that phosphene-based prosthetic vision could functionally recover some basic visual abilities (e.g., object perception and text reading) of non-congenitally blind individuals. Among these visual abilities, face perception is very important to humans due to its critical role in daily social interactions. Behavioral studies have tested and demonstrated the feasibility of face perception under phosphene-based prosthetic vision; however, neurophysiologic studies are required to elucidate the early visual processing mechanisms of phosphene faces.This thesis aimed to investigate the early face processing mechanism underlying the simulated prosthetic vision via electro-neurophysiologic and computational methods, which may not only provide an objective criterion for evaluating the efficacy of phosphene-based visual prostheses but also help test and improve the image processing strategies used in visual prosthesis design. The content of the thesis includes two main sections:In the first section, a face perception experiment was implemented using photographs of normal faces, non-face objects and their corresponding phosphene images. Electroencephalogram (EEG) was recorded with scalp electrodes during the experiment; and event-related potentials (ERPs) were derived by grand averaging the trials corresponding to each type of stimuli. We measured and analyzed face-sensitive ERP components at or near occipito-temporal areas (the ventral visual pathway), i.e., TP7/8, P7/8, PO7/8 and O1/2, to explore the early mechanism of face perception under prosthetic vision. Our results showed that (1) both normal and phosphene face stimuli could elicit prominent P1 and N170; (2) phosphene face stimuli caused significant amplitude suppression on N170 but not on P1 compared with normal stimuli; and (3) phosphene face stimuli also resulted in a significant delay on the latencies of both P1 and N170. Therefore, it was suggested that (1) early face processing was triggered in phosphene face perception; (2) the phosphene face pattern mainly impaired the fine processing stage rather than the coarse processing stage due to spatial discontinuity and information loss in high frequency; and (3) the phosphene pattern also disrupted the entire early processing of faces including both coarse and fine processing stages.In the second section, we employed independent component analysis (ICA) on the multi-channel EEGs that we recorded in the first section to seek for the independent face-sensitive EEG component. Then, we used source tracing algorithm and spherical four-shell BESA 2000 model to locate its source. Finally, we tested whether it contributed to face-sensitive ERP components P1 and N170 and whether it was modulated under phosphene faces. Our results showed that (1) an independent component IC-FS was found to be sensitive to face and the source of IC-FS located in fusiform gyrus; (2) IC-FS contributed to the face-sensitive P1 and N170 and related to face configural processing; (3) phosphene face pattern affected face configural processing so as the independent component IC-FS, leading to latency modulation on P1 and amplitude/latency modulation on N170.To sum up, this thesis systematically clarified how phosphene patterns affect the early face processing, offering a basic understanding of the neural representation of faces under phosphene vision. Meanwhile, it also provide a novel method for testing the efficacy of visual prosthesis, that is, evaluating the effect of prosthetic vision on face perception via recording electro-neurophysiologic data (EEG) and analysis the modulation of face sensitive ERPs and independent EEG components.