Best Frame Length determination for Magnetic Resonance Images Recognition

Abstract

Abstract Recently, the medical diagnosis via computers has become a hot research area. The division to frames is a vital process in image processing. It allows the examination of each part of the image. It is important to determine the ideal length of the frame because the division to very small frames increases the difficulty the recognizer. On the other hand, the division to big fi-ames makes many details lost. In this paper, the sequential search method is used to get the best frame length for the medical image recognition using artificial neural networks. Each image in the dataset is divided into 1,2, ..., ll frames and the features extracted from the fi'ames were used to train 11 neural networks. The best result was obtained when using 9 frames for an image of 200 pixels. This means that the optimal frame length for image recognition using neural networks is approximately 22 pixels for an image with any numbers of frames.