Elia Shahbazi¹, Drew Nguyen², Sophie Ledet¹, Rasel Ahmed Bhuiyan³, Adam Czajka³, Reza Azadi¹, Arash Afraz¹; ¹National Institutes of Health, ²University of Chicago, ³University of Notre Dame

We recently developed a paradigm called perceptography to visualize complex perceptual distortions induced by local brain stimulation. Perceptography uses machine-learning to create and optimize image distortions that are hard for the animal to distinguish from the state of being cortically stimulated. In our first project, we used optogenetics for brain-stimulation, which emphasized the concept but lacked simultaneous recording from stimulated neurons. Here and in our next move, using neural recording, we targeted specific groups of neurons within the inferior temporal (IT) cortex, including face selective neurons. We used two Utah-Array implants, one in the middle face-patch and one in the central IT-cortex. Macaque monkeys were trained to perform a cortical stimulation detection task, requiring fixation on a computer-generated image for 1s. At the midpoint of image presentation, we altered the image for 200ms and delivered an electrical stimulation impulse in 50% of the trials at random for 200ms. After the image, two targets appeared, and the animals indicated whether the stimulation occurred, earning liquid reward for correct reports. Our AI-generated framework, Perceptographer, enabled fine-grained perceptography for both face and non-face objects. Cortical stimulation in the face-patch produced perceptograms in which faces emerged for both face and object seed images, whereas stimulation outside the face-patch did not produce faces in object perceptograms. Furthermore, faces that emerged during face-patch stimulation appeared only contralateral to the stimulated hemisphere, while stimulation outside the face-patch produced whole-image distortions for both face and object seeds. These results indicate that representation of faces in IT-cortex is strongly contralateral, in contrast to broader bilateral representation of nonface objects. Neural recording corroborated these effects. Contralateral distortion/presentation elicited more vigorous neuronal activity in face neurons with face stimuli, whereas outside the face-patch, central bilateral image presentation/distortion elicited more robust responses for both face and object stimuli.