engleski

Neurocomputational model of superior visual search performance in autism spectrum disorder

We proposed a neural network model that explains visual search performance as a consequence of the formation of spatial map which labels surfaces according to their saliency. Labeling operation is achieved through the recurrent interactions which disinhibit all spatial locations occupied by the objects sharing the same feature values. The output of the network is further biased with feedback signals in order to split currently active nodes into a subset that satisfied more search constraints as it is required in the conjunction search. Furthermore, formation of the spatial map might be disrupted by the presence of neural noise in the input. Our suggestion is that autistic population is less susceptible to this kind of noise. Computer simulations showed that increased amount of noise fractionates the object’s representation in the spatial map which leads to the poorer visual search performance. In the conjunction search task, neural noise prevents selection of all objects sharing the same feature. Also, we showed that the same mechanism explain the search for embedded patterns which could underlie performance in embedded figures test.

autizam; pažnja; neuronski model

nije evidentirano