Talk by Dr. Yukako Yamane
Osaka University Graduate School of Frontier Biosciences, Japan
- 16 Nov 2012 13:30
- 16 Nov 2012 15:00
- INM-6, Bldg. 15.22, Seminar Room 3009, 1. OG
Representation of visual object shapes in monkey inferotemporal cortex
Inferotemporal (IT) cortex of macaque monkeys is the final visual area in the ventral visual pathway, which is important for object recognition. The neurons in this area respond to complex visual images, such as faces, hands and fractal patterns. Characterizing the responses of IT neurons is not so easy because we need to test many images within a limited recording time. I will talk about my two different experiments to characterize neuronal response of IT cortex efficiently. One is intrinsic signal optical imaging and the other is online adaptive stimulus creation.
1) By optical imaging, we visualized the location of highly active region (active spots) in the cortex for a particular visual stimulus. Then the electrode was inserted into observed active spots to determine the characteristics of each neuron in the active spot. This method not only visualizes the spread of active local area but makes it possible to focus on the neurons in the active spots and characterize its selectivity. An object image evoked multiple active spots in the IT cortex. Neurons in each spot responded to a part (not all) of an object image. This suggests that an object image is represented by multiple set of neurons each of which responds to a different part of the visual image.
2) We used an adaptive stimulus algorithm. A response-driven evolutionary algorithm guided initially random shapes through successive generations of morphed progeny, in effect letting the neuron design its own stimulus set from scratch. This procedure revealed that many IT neurons responded to any object containing a specific configuration of 3D surface fragments at specific object-relative 3D positions. These configurations corresponded to familiar geometric motifs like curving torsos and projecting limbs. They showed a strong bias toward high convex curvature, which may reflect the ecological importance of protruding object parts and our tendency to describe objects as arrangements of simple 3D parts. Our data indicate that, in addition to 2D image processing, the visual system generates an explicit representation of 3D object part structure.