DISCRIMINATION BETWEEN CHRONIC-PANCREATITIS AND PANCREATIC ADENOCARCINOMA USING ARTIFICIAL INTELLIGENCE-RELATED ALGORITHMS BASED ON IMAGE CYTOMETRY-GENERATED VARIABLES

The incidence of pancreatic adenocarcinomas (PA) is increased in the s etting of chronic pancreatitis. Distinguishing chronic pancreatitis ho rn pancreatic adenocarcinomas is often difficult, and is based on rout ine brush cytological specimens provided during endoscopic retrograde cholangiopancreatography (ERCP). Reactive epithelial changes in chroni c pancreatitis may appear similar to those of a well-differentiated ca ncer. Brush cytology specimens were obtained during ERCP from 49 patie nts with diseases for which the differential diagnosis included chroni c pancreatitis and/or pancreatic adenocarcinoma. Image cytometry was p erformed involving the assessment of between 200-400 Feulgen-stained n uclei per case; for each case, 40 quantitative cytometric variables we re generated. Data analysis was performed using artificial intelligenc e methods of data classification that produced decision trees and prod uction rule systems. Different classification models were produced for a subset of 34 patients, The best models were identified by the use o f a sampling technique (leave-one-out), and were tested on the remaini ng 15 patients. These models were based on 5 of the 40 variables assoc iated with a significant discriminatory function. Pancreatic adenocarc inoma was diagnosed in the training data set of 34 patients during a l eave-one-out process with an estimated sensitivity of 91% and specific ity of 87%. Both sensitivity and specificity were 80% in the independe nt test set of 15 patients. We conclude that inflammatory and malignan t pancreatic epithelia exhibit distinct morphological features that ca n be distinguished by decision tree-based classifiers employing image- cytometric numerical data. Cytometry 32:309-316, 1998. (C) 1998 Wiley- Liss, Inc.