This thesis presents a framework for the deployment of algorithms that support the correlation and real-time classification of physiological data streams through the development of clinically meaningful alerts using a blend of expert knowledge in the domain and pattern recognition programming based on clinical rules. Its relevance is demonstrated via a real world case study within the context of neonatal intensive care to provide real-time classification of neonatal spells. Events are first detected in individual streams independently; then synced together based on timestamps; and finally assessed to determine the start and end of a multi-signal episode. The episode is then processed through a classifier based on clinical rules to determine a classification. The output of the algorithms has been shown, in a single use case study with 24 hours of patient data, to detect clinically significant relative changes in heart rate, blood oxygen saturation levels and pauses in breathing in the respiratory impedance signal. The accuracy of the algorithm for detecting these is 97.8%, 98.3% and 98.9% respectively. The accuracy for correlating the streams and determining spells classifications is 98.9%. Future research will focus on the clinical validation of these algorithms and the application of the framework for the detection and classification of signals in other clinical contexts.