Classification of plasma cortisol patterns in normal subjects and in Cushing´s syndrome

A pattern recognition system for computer classification of cortisol time series into the normal class and subclasses of Cushing´s syndrome with different etiologies was developed. Discriminatory features are extracted from Fourier analysis and Karhunen-Loeve expansion coefficient or cortisol time series. Decision functions are trained by the least mean square error (LMSE) algorithm and tested by the jackknife test procedure on a database of 90 normal and patient patterns. The classification accuracy for normal, pituitary, adrenal, and ectopic classes is 100. 98.1, 98.3. and 100%, respectively. Hence, this pattern recognition system may be useful as an aid in the differential diagnosis of Cushing´s syndrome. Twenty-four-hour 24-h cortisol patterns can be easily obtained in a clinical research unit. This recognition system can be upgraded as new time series become available.