New Analytical Approaches in the Classification of Endocrine Diseases
Thursday, February 18, 1999 - 9:30am - 10:30am
Klaus Prank (Dartmouth Medical School)
In the search for better methods to define the temporal pattern of hormone secretion we and others have recently developed new tools to separate normal from diseased patterns of pulsatile hormone secretion and to reduce the amount of data necessary for such a classification. To classify the extracellular dynamics of hormonal fluctuations in the bloodstream as well as the dynamics of intracellular signaling pathways (e.g. [Ca2+]i -oscillations) we use nonlinear approaches such as artificial neural networks for time series prediction as well as approaches from information theory. These methods comprise the Approximate Entropy (ApEn) and Algorithmic Complexity (AC), measures for the regularity and complexity of a time series. These new approaches provide additional means for the classification of temporal patterns of secretion in health and disease when classical methods fail. Reducing the number of data points necessary to extract the important information for classifying the secretory dynamics of the hormone under study may in the future allow to use these analytical methods not only as research tools but also for routine diagnostic procedures.