Study Of Fusiform Gyrus Of Brain Region In Modeling Autism Using Artificial Neural Network

Abstract

In the present work, fusiform Gyrus (FG) of brain region has been studied, as many researchers have the opinion that autism is a neurological disorder of the brain. From various researchers, it has been observed that abnormalities in face perception are a core feature of social deficits in autism. The FG along with other cortical regions are responsible for face processing tasks in controls are hypoactive in patients with autism. Recently, through functional magnetic resonance imaging (fMRI), it has been observed that neurons in FG are fewer and smaller in patients with autism. In the present approach FG of brain region has been studied to analyze autism employing artificial neural network (ANN). As FG is responsible for face perception, the images of the whole faces are considered to be the input image. The images of faces are acquired and converted into gray scale images, which in turn, transformed into binary form. The produced binary information are fed into the backpropagation network (BPN) for learning. In this study the algorithm of backpropagation network has been implemented employing C++. The number of face images, size, learning rate, input and output neurons have been considered as 4, 50X50 pixels, 0.5, 2500, and 2 respectively, in the present BPN. At the time of train the network, the hidden neurons have been varied to observe the learning pattern of the FG for different set of input images. It has been observed that the learning pattern improved as hidden neurons increased for different input images. When neurons in the network increased, the face processing tasks of patients with autism seems to be normal. These results also justify the observation of fMRI for the autistic patients.

DOI: http://dx.doi.org/10.3329/jce.v26i1.10187

JCE 2011; 26(1): 66-69