Complex evaluation of secretin-pancreozymin test data by multivariate statistical pattern recognition methods

Abstract

Complex evaluation of secretin-pancreozymin test data with multivariate statistical methods was performed in patients with pancreatic insufficiency and in controls. The centroid, the nearest neighbour methods and the linear discriminant analysis led to the correct diagnosis in 84, 77 and 92%, respectively, when the overall responses to synthetic secretin plus cholecystokinin-octapeptide were evaluated. Results of individual stimulations were less precise. The linear discriminant analysis also successfully separated the patients with mild chronic pancreatitis from the healthy subjects, and seems to provide a useful support in medical diagnosis, particularly in controversial cases.