NeuroSerum: A Neural Net-based Diagnostic Aid Tool for Serum Electrophoresis

Using a standard laboratoty medical test, --serum electrophoresis--, several pathologic conditions, such as liver cirrhosis or nephrotic syndrome can be diagnosed. In all of these cases, the presence (in excess) or absence of certain types of proteins can be ascertained. This paper explains the status of the developement of an artificial neural net-based system that will help to diagnose these pathologies. We are going to use two different approximations to solve the problem. One is using the neural net structure Multi-Layer-Perceptron trained with the well known learning algorithm called BackPropagation. And the second is with another commonly used structure, Radial-Basis-Function also trained with a gradient descendent method. We try to reach an effective diagnostic aid system and to make a comparison of both types of nets working in classification tasks.