Neural Coding of Visual Stimuli

National Institute Of Mental Health

Investigators

Linked publications, trials & patents

Abstract

Visual perceptual categorization is a process whereby objects are assigned to groups based on similarity of visual features. This process requires both visual perception and visual memory. Two brain areas in the primate temporal lobe, areas TE and rhinal cortex are candidate regions for the neural substrates serving visual categorization. Area TE is considered to be the final stage of pure visual perception and object recognition. Rhinal cortex, the next stage of visual processing, is important as a node for visual memory. Whether or not rhinal cortex plays a role in visual perception is heavily debated. A previous selective lesion study from our lab showed that bilateral rhinal cortex lesions did not impair visual categorization, whereas bilateral TE removals mildly impaired visual categorization. Now we have investigated the role of TE and rhinal cortex in visual categorization further by making selective removals of each area and then testing the monkeys under conditions of increased task difficulty. We presume the monkeys need to accumulate perceptual evidence to respond correctly. To vary task difficulty, we manipulated the presentation duration of visual stimuli. Monkeys were trained to indicate whether a cue image, morphed between dogs and cats at 11 different levels of mixing, was more dog-like or cat-like. The cue image durations could be 25, 50, 100, 250 or 500ms, selected randomly trial-by-trial. Compared to unoperated controls, monkeys with bilateral TE lesions showed impaired categorization performance at all stimulus durations, but information processing time was not significantly different across the control and TE. Bilateral rhinal cortex lesions did not impair monkeys' categorization performance, but the information processing time was longer. That is, the monkeys with the rhinal cortex removals were slower to respond. For all groups, processing time increased with longer cue image durations and with increasing ambiguity of morph level. We also adopted a failure rate analysis to study the decision-making process. The results suggest that it takes more time for monkeys to make decisions after rhinal cortex lesions than it does for unoperated controls. TE removals did not affect the decision times. Overall, our results support the idea that TE might play a more perceptual role and rhinal cortex might play a more cognitive role in categorization-based decision making. One possibility is that rhinal cortex facilitates perceptual evidence accumulation via memory-guided attention. The classic hierarchical structure of visual processing (V1 -> V2 -> V4 -> TEO -> TE) posits that visual information reaches TE via TEO. A previous behavioral study from our lab suggests that visual categorization is performed by areas TEO and TE in parallel. These findings challenge the classic strictly sequential hierarchical structure of the ventral visual stream. Previous anatomical studies show that a direct projection from V4 to TE exists, independent of the V4-TEO-TE pathway. But it is unknown whether V4 neurons provide bifurcating afferent projections terminating in TEO and TE. In the present study, we injected nonreplicating viruses that are transported retrogradely from axon terminals throughout the neuron. We injected viruses expressing a colored fluorescent reporter into area TEO and a reporter with another color into TE. From these viral injections we can judge 1) the relative strengths of V4 projections to TEO and TE, 2) the general distribution of TEO-projecting and TE-projecting neurons in V4, and 3) the presence or absence of V4 neurons with bifurcated projections to both TEO and TE. To inject virus across TEO and TE, we used custom multichannel microinjector arrays. The high density of fluorescently labeled neurons in the targeted areas TEO and TE show that the viral delivery method worked well. We found populations of TEO-projecting neurons and TE-projecting neurons in V4, V4t and V4v, with intermixing of colored neurons in V4. In V4v, as has been seen in earlier studies. TEO-projecting neurons are in the more medial aspect, and TE-projecting neurons are in the more lateral aspect. In addition, there is a population of single neurons that project to both TEO and TE in V4, V4t and V4v as shown by co-localization of the two colors. Thus, in addition to the considerable projection from area V4 directly to area TE, a significant proportion of the information that reaches TE arises from this population that projects to both areas.