Humans recognize visual objects so effortlessly that is easy to overlook what an impressive computational feat this represents. At present, little is known about how the brain achieves robust visual object recognition, and reproducing this ability remains a major stumbling block in the construction of useful computer vision systems. In neurobiological research, rodent models have long been valued for their superior accessibility, with a wide range of powerful experimental techniques in widespread use. While rodents have not traditionally been used as a model system for object vision, recent behavioral evidence suggests that rats possess surprisingly advanced visual object recognition abilities. Building on these findings, the present project seeks to fill gaps in our knowledge of the neuronal mechanisms of vision using microelectrodes to record directly from neurons in the rat visual system. This work holds great potential to establish rodents as a new and powerful model for studying the neurophysiology of object recognition. The availability of a simpler, more accessible model can greatly accelerate progress in deciphering the computational underpinnings of high-level vision in the brain, which can, in turn, inform the construction of artificial vision systems for robotics and machine understanding of images. The project will also provide training opportunities for one postdoctoral fellow, and two undergraduate students, who will be directly involved in the data collection and analysis.
Humans have long sought to understand how the brain works. Because vision is one of our most important senses, much research has focused on understanding visual cortex -- the parts of the brain that are thought to enable us to identify objects and understand our visual environment. While significant progress has been made over the last fifty years, the mechanisms of vision are still not fully understood. To understand how visual cortex works, scientists rely heavily on animal models. The predominant animal model has been the non-human primate (monkey) because of its similarity to humans; however, while an important animal model, vision research in nonhuman primates is difficult and time-consuming to perform. In this project, we did work to establish rodents (specifically, rats) as an alternate/complementary animal model for research in high-level vision. We developed new research techniques that make it possible to use rodents to study high-level visual abilities, such as object recognition. We have also discovered neurons in a largely uncharted region of rat visual cortex that share properties with those found in non-human primates, suggesting that rodents are indeed a promising direction for future vision research. Establishing this new animal model promises to accelerate our understanding of visual cortex, because using rodents allows scientists to take advantage of the wide array of sophisticated tools that are already available for use with rodents (e.g., genetic manipulation). At the same time, rodents are inexpensive and easy to work with, and they provide a valuable experimental counterpoint to experiments in other species. Gaining a greater understanding of how visual cortex works promises not only to advance our understanding of ourselves, but also holds the potential for far reaching impact. In recent years, computational neuroscientists (including our laboratory) have built computer vision systems inspired by the structure and function of the brain, yielding impressive gains in commercially-relevant face and object recognition performance. The information we learn from researching simpler visual systems helps enable development of such computer vision models with broad potential impact.
