A new approach to building artificial neural networks and medicine

Abstract

Numerous attempts to use neural networks in medicine remain unsuccessful to this day because of an old mistake in the development of neural networks. In 1954, Frank Rosenblatt, created the first artificial neural network - the perceptron based on an understanding of the operation of brain neurons. This was a brilliant achievement; however, lack of knowledge at the time on how biological neurons worked led to systematic errors in the perceptron design and methods of training.  These errors have been repeatedly propagated in most artificial neural networks (ANN). The article describes the conceptual design of a brand-new type of perceptron named PANN (Progressive Artificial Neural Network), free from systematic errors of classical ANN and therefore with various unique properties. The article also provides data of the PANN network testing and a link that allows direct testing of the proposed neural network.

DOI: 10.5281/zenodo.5767079