Artificial neural networks application in medical images

Abstract

Diagnostic examinations through analog and numerical images are today the main method of diagnosing various diseases. Numerical images can be obtained with different methods depending on the technique used. Basically, any technique of obtaining medical images is nothing but the connection of a physical process of interaction of radiation with the subject / environment and with the help of the computer, these processes became visible through numerical images. Given the nature of the physical phenomena used and the lack of perfection of detection systems, the result is not perfect but is an estimate/price of “true value” which remains unattainable. Adding to this the human error during a medical examination, the patient's movements, etc., can flow very important artifacts, which must first be understood and analyzed and then using numerical methods, to correct them in the final version of the numerical image. This paper analyzes some numerical methods of numerical image correction such as interpolation and convolution, implemented in MATLAB program. In particular the interpolation technique is applied using Artificial Neural Networks (ANN), feedforward.