Abstract

This paper analyzes the behavior of Alzheimer’s disease simulations in an artificial neural network and based on the results proposes alternative possible diagnoses for Alzheimer’s disease. This is one of the most common fatal diseases, increasing in severity over time. Despite its high prevalence and thousands of yearly publications in this area, no cure has been found to date but in anticipation of a cure early detection is important towards fighting the disease.

The simulation of Alzheimer’s disease employs Hopfield memories. It is observed that the number of iterations needed to recognize distorted symbols is influenced by a small loss of connections while the recognition success rate stays surprisingly high for larger losses of elements. This is because the distortion enforces a search iterative process which is superfluous if the symbol tested is identical with the learned symbol. Hence, it is possible to suggest an early diagnostic approach which is based on recognizing e.g. characters of an alphabet with distorted or fragmented cues and measuring the time needed to perform the task, instead of merely measuring the subject’s success rate in the recognition process.

Keywords: Alzheimer´s disease; Early diagnosis; Distorted cues; Fragmented cues; Hopfield memories; Time test