An input classification scheme for use in evidence-based dynamic recurrent neuro-fuzzy prognosis.

This paper presents an input classification scheme used in an evidence-based dynamic recurrent neuro-fuzzy system for prognosis in rehabilitation. All external variables which may have an effect on the outcome of the rehabilitative process are classified into facts, contexts and interventions. Their effects on patients' physical and/or physiological states, which are estimated based on available evidence, are represented by fuzzy rules and/or non-linear models of physiologic processes. The outcomes of rehabilitation are defined as functions of those states.