Use of unsupervised neural networks for blood pressure profile classification

A methodology to classify blood pressure (BP) profiles with unsupervised learning neural networks is described. It can be used to discriminate different BP profile morphologies or hypertension levels (normotension, borderline, moderate and severe hypertension). After an extensive feasibility study, the Kohonen's topology preserving maps were chosen to identify similar morphologies in 100 BP profiles from different subjects. Afterwards, obtained results were validated using another group of 142 BP profiles.<<ETX>>