The invention relates to the modelling and design of early warning systems fordetecting medical conditions using physiological responses. The devicecomprisessensors for monitoring physiological parameters such as skin impedance, heartrate,and QT interval of a patient, means for establishing when those parameterschange,the rate of change of the parameters, and a neural network processor forprocessingthe information obtained by the sensors. The neural network processor isprogrammedwith a fast learning algorithm. When the neural network establishes that aphysiological condition is present in the patient an alarm signal will begenerated. Theinvention extends to a method of non-invasive monitoring of a person using aneuralnetwork programmed with a fast learning algorithm. A non-invasivehypoglycaemiamonitor is specifically described.
