An event-driven dynamic recurrent neuro-fuzzy system for adaptive prognosis in rehabilitation

An event-driven dynamic recurrent neuro-fuzzy system for prognosis in rehabilitation is introduced. Four layers are used to implement a fuzzy expert system, each of which consists of a cluster of neurons: input, rule-state, output, and outcome. The input "sensory" layer detects events based on physical and physiological status. Rules can fire in response to events and the current states, generating impulse or pulse-train signals to a given state. The states of the dynamic system, which often represent fuzzy expressions such as "strength.arm", change as a function of time. The event signals and the fuzzy states then are nonlinearly mapped to fuzzy outputs. Outcomes, similar to optimization performance criteria, are then a function of inputs, states and outputs. The system is designed with an interactive graphical user interface to facilitate rule creation by clinical experts. The structure is designed so that reinforcement learning approaches can be added in the future. This will enable tuning of system parameters, based on feedback from the patient and practitioner, as well as the error between predicted and measured outcomes.